Skip to contentSkip to conversation

2.6.32.9-stable review

PATCH In review Greg KH 8 participants 109 messages Feb 19, 2010 – Feb 26, 2010 1 message not in the archive

Subsystems (48)
lore
Share & tools
lore

Greg KH

author
#0 lore raw

cover 2.6.32.9-stable review

This is the start of the stable review cycle for the 2.6.32.9 release. There are 93 patches in this series, all will be posted as a response to this one. If anyone has any issues with these being applied, please let us know. If anyone is a maintainer of the proper subsystem, and wants to add a Signed-off-by: line to the patch, please respond with it.

There are still a number of patches that I have had submitted for inclusion in the .32 kernel series that I have not gotten to yet. This release has taken longer than I wanted (again), and the size of it, and the importance of it, justifies pushing it out now. If you have sent patches for inclusion that you do not see here, please wait for the 2.6.32.10 release before worrying.

Responses should be made by Sunday, February 21, 2010, 17:00:00 UTC. Anything received after that time might be too late.

The whole patch series can be found in one patch at:

	kernel.org/pub/linux/kernel/v2.6/stable-review/patch-2.6.32.9-rc1.gz

and the diffstat can be found below.

thanks,

greg k-h

 Makefile                                           |    2 +-
 arch/sh/include/asm/pgtable_32.h                   |    5 +-
 arch/x86/include/asm/amd_iommu.h                   |    1 +
 arch/x86/kernel/amd_iommu.c                        |   12 ++-
 arch/x86/kernel/amd_iommu_init.c                   |    3 +
 arch/x86/kernel/apic/io_apic.c                     |    1 +
 arch/x86/kernel/cpu/cpufreq/powernow-k8.c          |    3 +-
 arch/x86/kvm/i8254.c                               |    3 +
 arch/x86/kvm/x86.c                                 |    7 +-
 drivers/acpi/processor_idle.c                      |    8 ++
 drivers/acpi/scan.c                                |    3 +
 drivers/ata/ahci.c                                 |   15 +++
 drivers/ata/libata-sff.c                           |    3 +
 drivers/base/class.c                               |    2 +
 drivers/block/cciss.c                              |    3 +
 drivers/char/mem.c                                 |   70 ++++++------
 drivers/char/tpm/tpm_infineon.c                    |   79 +++++++++----
 drivers/char/tty_io.c                              |    4 +-
 drivers/crypto/padlock-sha.c                       |   23 ++++
 drivers/dma/ioat/dma_v2.c                          |    2 +-
 drivers/edac/amd64_edac.c                          |   15 ++-
 drivers/gpu/drm/i915/i915_debugfs.c                |   30 -----
 drivers/gpu/drm/i915/i915_drv.h                    |   13 ++
 drivers/gpu/drm/i915/i915_gem.c                    |  123 +++++++++++++++++---
 drivers/gpu/drm/i915/i915_reg.h                    |   15 +++
 drivers/gpu/drm/i915/i915_suspend.c                |    5 +-
 drivers/gpu/drm/i915/intel_crt.c                   |    3 +
 drivers/gpu/drm/i915/intel_display.c               |   48 +++++++-
 drivers/gpu/drm/i915/intel_drv.h                   |    2 +
 drivers/gpu/drm/i915/intel_fb.c                    |    2 +-
 drivers/gpu/drm/i915/intel_i2c.c                   |   19 +++
 drivers/gpu/drm/i915/intel_lvds.c                  |   21 ++++
 drivers/gpu/drm/radeon/atom.c                      |    2 +-
 drivers/hwmon/adt7462.c                            |    2 +-
 drivers/hwmon/lm78.c                               |   25 ++--
 drivers/hwmon/w83781d.c                            |   26 ++--
 drivers/i2c/busses/i2c-tiny-usb.c                  |   12 +-
 drivers/md/dm-log-userspace-transfer.c             |   10 +-
 drivers/md/dm-stripe.c                             |    2 +-
 drivers/md/dm.c                                    |   21 +++-
 drivers/md/raid5.c                                 |   11 +-
 drivers/media/dvb/dvb-core/dmxdev.c                |    2 +-
 drivers/message/fusion/mptscsih.c                  |    2 +-
 drivers/net/wireless/ath/ath9k/xmit.c              |    4 +-
 drivers/net/wireless/b43/b43.h                     |    1 +
 drivers/net/wireless/b43/main.c                    |   13 ++-
 drivers/net/wireless/iwlwifi/iwl-core.c            |    1 +
 drivers/net/wireless/rtl818x/rtl8187_dev.c         |    1 +
 drivers/platform/x86/dell-wmi.c                    |    8 +-
 drivers/platform/x86/hp-wmi.c                      |    9 ++-
 drivers/platform/x86/wmi.c                         |    4 +-
 drivers/regulator/core.c                           |    2 +-
 drivers/rtc/rtc-fm3130.c                           |    6 +-
 drivers/s390/block/dasd.c                          |   22 ++--
 drivers/s390/block/dasd_eckd.c                     |   44 +++----
 drivers/s390/block/dasd_fba.c                      |   10 +-
 drivers/s390/block/dasd_int.h                      |   10 ++
 drivers/serial/8250.c                              |    7 +-
 drivers/staging/rtl8187se/ieee80211/ieee80211.h    |   10 +-
 drivers/staging/rtl8187se/ieee80211/ieee80211_rx.c |    2 +-
 .../rtl8187se/ieee80211/ieee80211_softmac.c        |   14 +-
 drivers/staging/rtl8187se/ieee80211/ieee80211_tx.c |    2 +-
 drivers/staging/rtl8187se/r8180_core.c             |   10 +-
 drivers/staging/rtl8187se/r8180_wx.c               |    2 +-
 drivers/usb/core/devio.c                           |   48 +++++---
 drivers/usb/host/r8a66597-hcd.c                    |   37 ++++--
 fs/befs/linuxvfs.c                                 |    1 +
 fs/block_dev.c                                     |    7 +-
 fs/cifs/readdir.c                                  |    1 +
 fs/ecryptfs/inode.c                                |   16 +++
 fs/exec.c                                          |   21 +++-
 fs/fcntl.c                                         |    6 +-
 fs/nfs/direct.c                                    |    3 +
 fs/nfs/file.c                                      |    2 +
 fs/nfs/fscache.c                                   |    9 +-
 fs/nfs/mount_clnt.c                                |    2 +-
 fs/nfs/nfs2xdr.c                                   |    2 +-
 fs/nfs/nfs4_fs.h                                   |    1 +
 fs/nfs/nfs4proc.c                                  |   26 ++++-
 fs/nfs/nfs4xdr.c                                   |    6 +-
 fs/nfs/pagelist.c                                  |   17 ++--
 fs/nfs/super.c                                     |   15 +++-
 fs/nfs/write.c                                     |    1 +
 fs/sysfs/inode.c                                   |   47 ++++----
 include/drm/drm_os_linux.h                         |    2 +-
 include/linux/sched.h                              |   22 ++++
 include/net/netns/conntrack.h                      |    3 +
 include/net/netns/ipv4.h                           |    1 +
 kernel/futex.c                                     |   30 +++++-
 kernel/time/timekeeping.c                          |    2 +
 mm/migrate.c                                       |    3 +
 net/core/dst.c                                     |    2 +
 net/core/pktgen.c                                  |    1 +
 net/ipv4/netfilter/arp_tables.c                    |    4 +-
 net/ipv4/netfilter/ip_tables.c                     |    4 +-
 net/ipv4/netfilter/nf_conntrack_l3proto_ipv4.c     |    2 +-
 .../netfilter/nf_conntrack_l3proto_ipv4_compat.c   |    4 +-
 net/ipv4/netfilter/nf_nat_core.c                   |   22 ++--
 net/ipv6/netfilter/ip6_tables.c                    |    4 +-
 net/mac80211/ibss.c                                |    2 +-
 net/netfilter/nf_conntrack_core.c                  |  117 ++++++++++---------
 net/netfilter/nf_conntrack_expect.c                |    4 +-
 net/netfilter/nf_conntrack_helper.c                |    2 +-
 net/netfilter/nf_conntrack_netlink.c               |    2 +-
 net/netfilter/nf_conntrack_standalone.c            |    7 +-
 sound/pci/ctxfi/ctatc.c                            |   15 +--
 sound/pci/ctxfi/ctvmem.c                           |   38 +++---
 sound/pci/ctxfi/ctvmem.h                           |    8 +-
 sound/pci/hda/hda_intel.c                          |    3 +
 sound/pci/hda/patch_realtek.c                      |   28 ++++-
 sound/usb/usbaudio.c                               |    2 +-
 111 files changed, 981 insertions(+), 463 deletions(-)
author
#1 lore raw

PATCH 1/93 Fix potential crash with sys_move_pages

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Linus Torvalds <[email protected]>

commit 6f5a55f1a6c5abee15a0e878e5c74d9f1569b8b0 upstream.

We incorrectly depended on the 'node_state/node_isset()' functions testing the node range, rather than checking it explicitly. That's not reliable, even if it might often happen to work. So do the proper explicit test.

---
 mm/migrate.c |    3 +++
 1 file changed, 3 insertions(+)
mm/migrate.c +3 -0
@@ -953,6 +953,9 @@ static int do_pages_move(struct mm_struc
 				goto out_pm;
 
 			err = -ENODEV;
+			if (node < 0 || node >= MAX_NUMNODES)
+				goto out_pm;
+
 			if (!node_state(node, N_HIGH_MEMORY))
 				goto out_pm;
 
author
#2 lore raw

PATCH 2/93 futex_lock_pi() key refcnt fix

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Mikael Pettersson <[email protected]>

commit 5ecb01cfdf96c5f465192bdb2a4fd4a61a24c6cc upstream.

This fixes a futex key reference count bug in futex_lock_pi(), where a key's reference count is incremented twice but decremented only once, causing the backing object to not be released.

If the futex is created in a temporary file in an ext3 file system, this bug causes the file's inode to become an "undead" orphan, which causes an oops from a BUG_ON() in ext3_put_super() when the file system is unmounted. glibc's test suite is known to trigger this, see <http://bugzilla.kernel.org/show_bug.cgi?id=14256>.

The bug is a regression from 2.6.28-git3, namely Peter Zijlstra's 38d47c1b7075bd7ec3881141bb3629da58f88dab "[PATCH] futex: rely on get_user_pages() for shared futexes". That commit made get_futex_key() also increment the reference count of the futex key, and updated its callers to decrement the key's reference count before returning. Unfortunately the normal exit path in futex_lock_pi() wasn't corrected:
the reference count is incremented by get_futex_key() and queue_lock(), but the normal exit path only decrements once, via unqueue_me_pi(). The fix is to put_futex_key() after unqueue_me_pi(), since 2.6.31 this is easily done by 'goto out_put_key' rather than 'goto out'.

---
 kernel/futex.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
kernel/futex.c +1 -1
@@ -1971,7 +1971,7 @@ retry_private:
 	/* Unqueue and drop the lock */
 	unqueue_me_pi(&q);
 
-	goto out;
+	goto out_put_key;
 
 out_unlock_put_key:
 	queue_unlock(&q, hb);
author
#3 lore raw

PATCH 3/93 futex: Handle user space corruption gracefully

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Thomas Gleixner <[email protected]>

commit 51246bfd189064079c54421507236fd2723b18f3 upstream.

If the owner of a PI futex dies we fix up the pi_state and set pi_state->owner to NULL. When a malicious or just sloppy programmed user space application sets the futex value to 0 e.g. by calling pthread_mutex_init(), then the futex can be acquired again. A new waiter manages to enqueue itself on the pi_state w/o damage, but on unlock the kernel dereferences pi_state->owner and oopses.

Prevent this by checking pi_state->owner in the unlock path. If pi_state->owner is not current we know that user space manipulated the futex value. Ignore the mess and return -EINVAL.

This catches the above case and also the case where a task hijacks the futex by setting the tid value and then tries to unlock it.

---
 kernel/futex.c |    7 +++++++
 1 file changed, 7 insertions(+)
kernel/futex.c +7 -0
@@ -758,6 +758,13 @@ static int wake_futex_pi(u32 __user *uad
 	if (!pi_state)
 		return -EINVAL;
 
+	/*
+	 * If current does not own the pi_state then the futex is
+	 * inconsistent and user space fiddled with the futex value.
+	 */
+	if (pi_state->owner != current)
+		return -EINVAL;
+
 	spin_lock(&pi_state->pi_mutex.wait_lock);
 	new_owner = rt_mutex_next_owner(&pi_state->pi_mutex);
 
author
#4 lore raw

PATCH 4/93 futex: Handle futex value corruption gracefully

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Thomas Gleixner <[email protected]>

commit 59647b6ac3050dd964bc556fe6ef22f4db5b935c upstream.

The WARN_ON in lookup_pi_state which complains about a mismatch between pi_state->owner->pid and the pid which we retrieved from the user space futex is completely bogus.

The code just emits the warning and then continues despite the fact that it detected an inconsistent state of the futex. A conveniant way for user space to spam the syslog.

Replace the WARN_ON by a consistency check. If the values do not match return -EINVAL and let user space deal with the mess it created.

This also fixes the missing task_pid_vnr() when we compare the pi_state->owner pid with the futex value.

---
 kernel/futex.c |   21 +++++++++++++++++++--
 1 file changed, 19 insertions(+), 2 deletions(-)
kernel/futex.c +19 -2
@@ -530,8 +530,25 @@ lookup_pi_state(u32 uval, struct futex_h
 				return -EINVAL;
 
 			WARN_ON(!atomic_read(&pi_state->refcount));
-			WARN_ON(pid && pi_state->owner &&
-				pi_state->owner->pid != pid);
+
+			/*
+			 * When pi_state->owner is NULL then the owner died
+			 * and another waiter is on the fly. pi_state->owner
+			 * is fixed up by the task which acquires
+			 * pi_state->rt_mutex.
+			 *
+			 * We do not check for pid == 0 which can happen when
+			 * the owner died and robust_list_exit() cleared the
+			 * TID.
+			 */
+			if (pid && pi_state->owner) {
+				/*
+				 * Bail out if user space manipulated the
+				 * futex value.
+				 */
+				if (pid != task_pid_vnr(pi_state->owner))
+					return -EINVAL;
+			}
 
 			atomic_inc(&pi_state->refcount);
 			*ps = pi_state;
author
#5 lore raw

PATCH 5/93 Fix race in tty_fasync() properly

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Linus Torvalds <[email protected]>

commit 80e1e823989ec44d8e35bdfddadbddcffec90424 upstream.

This reverts commit 703625118069 ("tty: fix race in tty_fasync") and commit b04da8bfdfbb ("fnctl: f_modown should call write_lock_irqsave/ restore") that tried to fix up some of the fallout but was incomplete.

It turns out that we really cannot hold 'tty->ctrl_lock' over calling __f_setown, because not only did that cause problems with interrupt disables (which the second commit fixed), it also causes a potential ABBA deadlock due to lock ordering.

Thanks to Tetsuo Handa for following up on the issue, and running lockdep to show the problem. It goes roughly like this:

- f_getown gets filp->f_owner.lock for reading without interrupts disabled, so an interrupt that happens while that lock is held can cause a lockdep chain from f_owner.lock -> sighand->siglock.

- at the same time, the tty->ctrl_lock -> f_owner.lock chain that commit 703625118069 introduced, together with the pre-existing sighand->siglock -> tty->ctrl_lock chain means that we have a lock dependency the other way too.

So instead of extending tty->ctrl_lock over the whole __f_setown() call, we now just take a reference to the 'pid' structure while holding the lock, and then release it after having done the __f_setown. That still guarantees that 'struct pid' won't go away from under us, which is all we really ever needed.

---
 drivers/char/tty_io.c |    4 +++-
 fs/fcntl.c            |    6 ++----
 2 files changed, 5 insertions(+), 5 deletions(-)
drivers/char/tty_io.c +3 -1
@@ -1930,8 +1930,10 @@ static int tty_fasync(int fd, struct fil
 			pid = task_pid(current);
 			type = PIDTYPE_PID;
 		}
-		retval = __f_setown(filp, pid, type, 0);
+		get_pid(pid);
 		spin_unlock_irqrestore(&tty->ctrl_lock, flags);
+		retval = __f_setown(filp, pid, type, 0);
+		put_pid(pid);
 		if (retval)
 			goto out;
 	} else {
fs/fcntl.c +2 -4
@@ -199,9 +199,7 @@ static int setfl(int fd, struct file * f
 static void f_modown(struct file *filp, struct pid *pid, enum pid_type type,
                      int force)
 {
-	unsigned long flags;
-
-	write_lock_irqsave(&filp->f_owner.lock, flags);
+	write_lock_irq(&filp->f_owner.lock);
 	if (force || !filp->f_owner.pid) {
 		put_pid(filp->f_owner.pid);
 		filp->f_owner.pid = get_pid(pid);
@@ -213,7 +211,7 @@ static void f_modown(struct file *filp,
 			filp->f_owner.euid = cred->euid;
 		}
 	}
-	write_unlock_irqrestore(&filp->f_owner.lock, flags);
+	write_unlock_irq(&filp->f_owner.lock);
 }
 
 int __f_setown(struct file *filp, struct pid *pid, enum pid_type type,
author
#6 lore raw

PATCH 6/93 hwmon: (w83781d) Request I/O ports individually for probing

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Jean Delvare <[email protected]>

commit b0bcdd3cd0adb85a7686b396ba50493871b1135c upstream.

Different motherboards have different PNP declarations for W83781D/W83782D chips. Some declare the whole range of I/O ports (8 ports), some declare only the useful ports (2 ports at offset 5) and some declare fancy ranges, for example 4 ports at offset 4. To properly handle all cases, request all ports individually for probing. After we have determined that we really have a W83781D or W83782D chip, the useful port range will be requested again, as a single block.

I did not see a board which needs this yet, but I know of one for lm78 driver and I'd like to keep the logic of these two drivers in sync.

---
 drivers/hwmon/w83781d.c |   24 ++++++++++++------------
 1 file changed, 12 insertions(+), 12 deletions(-)
drivers/hwmon/w83781d.c +12 -12
@@ -1818,17 +1818,17 @@ static int __init
 w83781d_isa_found(unsigned short address)
 {
 	int val, save, found = 0;
+	int port;
 
-	/* We have to request the region in two parts because some
-	   boards declare base+4 to base+7 as a PNP device */
-	if (!request_region(address, 4, "w83781d")) {
-		pr_debug("w83781d: Failed to request low part of region\n");
-		return 0;
-	}
-	if (!request_region(address + 4, 4, "w83781d")) {
-		pr_debug("w83781d: Failed to request high part of region\n");
-		release_region(address, 4);
-		return 0;
+	/* Some boards declare base+0 to base+7 as a PNP device, some base+4
+	 * to base+7 and some base+5 to base+6. So we better request each port
+	 * individually for the probing phase. */
+	for (port = address; port < address + W83781D_EXTENT; port++) {
+		if (!request_region(port, 1, "w83781d")) {
+			pr_debug("w83781d: Failed to request port 0x%x\n",
+				 port);
+			goto release;
+		}
 	}
 
 #define REALLY_SLOW_IO
@@ -1902,8 +1902,8 @@ w83781d_isa_found(unsigned short address
 			val == 0x30 ? "W83782D" : "W83781D", (int)address);
 
  release:
-	release_region(address + 4, 4);
-	release_region(address, 4);
+	for (port--; port >= address; port--)
+		release_region(port, 1);
 	return found;
 }
 
author
#7 lore raw

PATCH 7/93 hwmon: (lm78) Request I/O ports individually for probing

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Jean Delvare <[email protected]>

commit 197027e6ef830d60e10f76efc8d12bf3b6c35db5 upstream.

Different motherboards have different PNP declarations for LM78/LM79 chips. Some declare the whole range of I/O ports (8 ports), some declare only the useful ports (2 ports at offset 5) and some declare fancy ranges, for example 4 ports at offset 4. To properly handle all cases, request all ports individually for probing. After we have determined that we really have an LM78 or LM79 chip, the useful port range will be requested again, as a single block.

This fixes the driver on the Olivetti M3000 DT 540, at least.

---
 drivers/hwmon/lm78.c |   23 +++++++++++------------
 1 file changed, 11 insertions(+), 12 deletions(-)
drivers/hwmon/lm78.c +11 -12
@@ -870,17 +870,16 @@ static struct lm78_data *lm78_update_dev
 static int __init lm78_isa_found(unsigned short address)
 {
 	int val, save, found = 0;
+	int port;
 
-	/* We have to request the region in two parts because some
-	   boards declare base+4 to base+7 as a PNP device */
-	if (!request_region(address, 4, "lm78")) {
-		pr_debug("lm78: Failed to request low part of region\n");
-		return 0;
-	}
-	if (!request_region(address + 4, 4, "lm78")) {
-		pr_debug("lm78: Failed to request high part of region\n");
-		release_region(address, 4);
-		return 0;
+	/* Some boards declare base+0 to base+7 as a PNP device, some base+4
+	 * to base+7 and some base+5 to base+6. So we better request each port
+	 * individually for the probing phase. */
+	for (port = address; port < address + LM78_EXTENT; port++) {
+		if (!request_region(port, 1, "lm78")) {
+			pr_debug("lm78: Failed to request port 0x%x\n", port);
+			goto release;
+		}
 	}
 
 #define REALLY_SLOW_IO
@@ -944,8 +943,8 @@ static int __init lm78_isa_found(unsigne
 			val & 0x80 ? "LM79" : "LM78", (int)address);
 
  release:
-	release_region(address + 4, 4);
-	release_region(address, 4);
+	for (port--; port >= address; port--)
+		release_region(port, 1);
 	return found;
 }
 
author
#8 lore raw

PATCH 8/93 hwmon: (adt7462) Wrong ADT7462_VOLT_COUNT

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Ray Copeland <[email protected]>

commit 85f8d3e5faea8bd36c3e5196f8334f7db45e19b2 upstream.

The #define ADT7462_VOLT_COUNT is wrong, it should be 13 not 12. All the for loops that use this as a limit count are of the typical form, "for (n = 0; n < ADT7462_VOLT_COUNT; n++)", so to loop through all voltages w/o missing the last one it is necessary for the count to be one greater than it is. (Specifically, you will miss the +1.5V 3GPIO input with count = 12 vs. 13.)

---
 drivers/hwmon/adt7462.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/hwmon/adt7462.c +1 -1
@@ -182,7 +182,7 @@ I2C_CLIENT_INSMOD_1(adt7462);
  *
  * Some, but not all, of these voltages have low/high limits.
  */
-#define ADT7462_VOLT_COUNT	12
+#define ADT7462_VOLT_COUNT	13
 
 #define ADT7462_VENDOR		0x41
 #define ADT7462_DEVICE		0x62
author
#9 lore raw

PATCH 9/93 ALSA: ctxfi - fix PTP address initialization

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Jaroslav Kysela <[email protected]>

commit 21956b61f594f7924d98240da74bc81c28601fa9 upstream.

After hours of debugging, I finally found the reason why some source and runtime combination does not work. The PTP (page table pages) address must be aligned. I am not sure how much, but alignment to PAGE_SIZE is sufficient. Also, use ALSA's page allocation routines to ensure proper virtual -> physical address translation.

---
 sound/pci/ctxfi/ctatc.c  |   15 ++-------------
 sound/pci/ctxfi/ctvmem.c |   38 ++++++++++++++++++--------------------
 sound/pci/ctxfi/ctvmem.h |    8 +++++---
 3 files changed, 25 insertions(+), 36 deletions(-)
sound/pci/ctxfi/ctatc.c +2 -13
@@ -166,18 +166,7 @@ static void ct_unmap_audio_buffer(struct
 
 static unsigned long atc_get_ptp_phys(struct ct_atc *atc, int index)
 {
-	struct ct_vm *vm;
-	void *kvirt_addr;
-	unsigned long phys_addr;
-
-	vm = atc->vm;
-	kvirt_addr = vm->get_ptp_virt(vm, index);
-	if (kvirt_addr == NULL)
-		phys_addr = (~0UL);
-	else
-		phys_addr = virt_to_phys(kvirt_addr);
-
-	return phys_addr;
+	return atc->vm->get_ptp_phys(atc->vm, index);
 }
 
 static unsigned int convert_format(snd_pcm_format_t snd_format)
@@ -1669,7 +1658,7 @@ int __devinit ct_atc_create(struct snd_c
 	}
 
 	/* Set up device virtual memory management object */
-	err = ct_vm_create(&atc->vm);
+	err = ct_vm_create(&atc->vm, pci);
 	if (err < 0)
 		goto error1;
 
sound/pci/ctxfi/ctvmem.c +18 -20
@@ -138,7 +138,7 @@ ct_vm_map(struct ct_vm *vm, struct snd_p
 		return NULL;
 	}
 
-	ptp = vm->ptp[0];
+	ptp = (unsigned long *)vm->ptp[0].area;
 	pte_start = (block->addr >> CT_PAGE_SHIFT);
 	pages = block->size >> CT_PAGE_SHIFT;
 	for (i = 0; i < pages; i++) {
@@ -158,25 +158,25 @@ static void ct_vm_unmap(struct ct_vm *vm
 }
 
 /* *
- * return the host (kmalloced) addr of the @index-th device
- * page talbe page on success, or NULL on failure.
- * The first returned NULL indicates the termination.
+ * return the host physical addr of the @index-th device
+ * page table page on success, or ~0UL on failure.
+ * The first returned ~0UL indicates the termination.
  * */
-static void *
-ct_get_ptp_virt(struct ct_vm *vm, int index)
+static dma_addr_t
+ct_get_ptp_phys(struct ct_vm *vm, int index)
 {
-	void *addr;
+	dma_addr_t addr;
 
-	addr = (index >= CT_PTP_NUM) ? NULL : vm->ptp[index];
+	addr = (index >= CT_PTP_NUM) ? ~0UL : vm->ptp[index].addr;
 
 	return addr;
 }
 
-int ct_vm_create(struct ct_vm **rvm)
+int ct_vm_create(struct ct_vm **rvm, struct pci_dev *pci)
 {
 	struct ct_vm *vm;
 	struct ct_vm_block *block;
-	int i;
+	int i, err = 0;
 
 	*rvm = NULL;
 
@@ -188,23 +188,21 @@ int ct_vm_create(struct ct_vm **rvm)
 
 	/* Allocate page table pages */
 	for (i = 0; i < CT_PTP_NUM; i++) {
-		vm->ptp[i] = kmalloc(PAGE_SIZE, GFP_KERNEL);
-		if (!vm->ptp[i])
+		err = snd_dma_alloc_pages(SNDRV_DMA_TYPE_DEV,
+					  snd_dma_pci_data(pci),
+					  PAGE_SIZE, &vm->ptp[i]);
+		if (err < 0)
 			break;
 	}
-	if (!i) {
+	if (err < 0) {
 		/* no page table pages are allocated */
-		kfree(vm);
+		ct_vm_destroy(vm);
 		return -ENOMEM;
 	}
 	vm->size = CT_ADDRS_PER_PAGE * i;
-	/* Initialise remaining ptps */
-	for (; i < CT_PTP_NUM; i++)
-		vm->ptp[i] = NULL;
-
 	vm->map = ct_vm_map;
 	vm->unmap = ct_vm_unmap;
-	vm->get_ptp_virt = ct_get_ptp_virt;
+	vm->get_ptp_phys = ct_get_ptp_phys;
 	INIT_LIST_HEAD(&vm->unused);
 	INIT_LIST_HEAD(&vm->used);
 	block = kzalloc(sizeof(*block), GFP_KERNEL);
@@ -242,7 +240,7 @@ void ct_vm_destroy(struct ct_vm *vm)
 
 	/* free allocated page table pages */
 	for (i = 0; i < CT_PTP_NUM; i++)
-		kfree(vm->ptp[i]);
+		snd_dma_free_pages(&vm->ptp[i]);
 
 	vm->size = 0;
 
sound/pci/ctxfi/ctvmem.h +5 -3
@@ -22,6 +22,8 @@
 
 #include <linux/mutex.h>
 #include <linux/list.h>
+#include <linux/pci.h>
+#include <sound/memalloc.h>
 
 /* The chip can handle the page table of 4k pages
  * (emu20k1 can handle even 8k pages, but we don't use it right now)
@@ -41,7 +43,7 @@ struct snd_pcm_substream;
 
 /* Virtual memory management object for card device */
 struct ct_vm {
-	void *ptp[CT_PTP_NUM];		/* Device page table pages */
+	struct snd_dma_buffer ptp[CT_PTP_NUM];	/* Device page table pages */
 	unsigned int size;		/* Available addr space in bytes */
 	struct list_head unused;	/* List of unused blocks */
 	struct list_head used;		/* List of used blocks */
@@ -52,10 +54,10 @@ struct ct_vm {
 				   int size);
 	/* Unmap device logical addr area. */
 	void (*unmap)(struct ct_vm *, struct ct_vm_block *block);
-	void *(*get_ptp_virt)(struct ct_vm *vm, int index);
+	dma_addr_t (*get_ptp_phys)(struct ct_vm *vm, int index);
 };
 
-int ct_vm_create(struct ct_vm **rvm);
+int ct_vm_create(struct ct_vm **rvm, struct pci_dev *pci);
 void ct_vm_destroy(struct ct_vm *vm);
 
 #endif /* CTVMEM_H */
author
#10 lore raw

PATCH 10/93 drm/i915: disable hotplug detect before Ironlake CRT detect

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Zhenyu Wang <[email protected]>

commit eceb784cec4dc0fcc2993d9ee4a7c0d111ada80a upstream.

This tries to fix CRT detect loop hang seen on some Ironlake form factor, to clear up hotplug detect state before taking CRT detect to make sure next hotplug detect cycle is consistent.

---
 drivers/gpu/drm/i915/intel_crt.c |    3 +++
 1 file changed, 3 insertions(+)
drivers/gpu/drm/i915/intel_crt.c +3 -0
@@ -185,6 +185,9 @@ static bool intel_igdng_crt_detect_hotpl
 	adpa = I915_READ(PCH_ADPA);
 
 	adpa &= ~ADPA_CRT_HOTPLUG_MASK;
+	/* disable HPD first */
+	I915_WRITE(PCH_ADPA, adpa);
+	(void)I915_READ(PCH_ADPA);
 
 	adpa |= (ADPA_CRT_HOTPLUG_PERIOD_128 |
 			ADPA_CRT_HOTPLUG_WARMUP_10MS |
author
#11 lore raw

PATCH 11/93 drm/i915: enable self-refresh on 965

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Jesse Barnes <jbarnes@jbarnes-x200.(none)>

commit 1dc7546d1a73664e5d117715b214bea9cae5951c upstream.

Need to calculate the SR watermark and enable it.

---
 drivers/gpu/drm/i915/intel_display.c |   32 ++++++++++++++++++++++++++++----
 1 file changed, 28 insertions(+), 4 deletions(-)
drivers/gpu/drm/i915/intel_display.c +28 -4
@@ -2556,15 +2556,39 @@ static void g4x_update_wm(struct drm_dev
 		   (cursor_sr << DSPFW_CURSOR_SR_SHIFT));
 }
 
-static void i965_update_wm(struct drm_device *dev, int unused, int unused2,
-			   int unused3, int unused4)
+static void i965_update_wm(struct drm_device *dev, int planea_clock,
+			   int planeb_clock, int sr_hdisplay, int pixel_size)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
+	unsigned long line_time_us;
+	int sr_clock, sr_entries, srwm = 1;
 
-	DRM_DEBUG("Setting FIFO watermarks - A: 8, B: 8, C: 8, SR 8\n");
+	/* Calc sr entries for one plane configs */
+	if (sr_hdisplay && (!planea_clock || !planeb_clock)) {
+		/* self-refresh has much higher latency */
+		const static int sr_latency_ns = 12000;
+
+		sr_clock = planea_clock ? planea_clock : planeb_clock;
+		line_time_us = ((sr_hdisplay * 1000) / sr_clock);
+
+		/* Use ns/us then divide to preserve precision */
+		sr_entries = (((sr_latency_ns / line_time_us) + 1) *
+			      pixel_size * sr_hdisplay) / 1000;
+		sr_entries = roundup(sr_entries / I915_FIFO_LINE_SIZE, 1);
+		DRM_DEBUG("self-refresh entries: %d\n", sr_entries);
+		srwm = I945_FIFO_SIZE - sr_entries;
+		if (srwm < 0)
+			srwm = 1;
+		srwm &= 0x3f;
+		I915_WRITE(FW_BLC_SELF, FW_BLC_SELF_EN);
+	}
+
+	DRM_DEBUG_KMS("Setting FIFO watermarks - A: 8, B: 8, C: 8, SR %d\n",
+		      srwm);
 
 	/* 965 has limitations... */
-	I915_WRITE(DSPFW1, (8 << 16) | (8 << 8) | (8 << 0));
+	I915_WRITE(DSPFW1, (srwm << DSPFW_SR_SHIFT) | (8 << 16) | (8 << 8) |
+		   (8 << 0));
 	I915_WRITE(DSPFW2, (8 << 8) | (8 << 0));
 }
 
author
#12 lore raw

PATCH 12/93 drm/i915: Disable SR when more than one pipe is enabled

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: David John <[email protected]>

commit 33c5fd121eabbccc9103daf6cda36941eb3c349f upstream.

Self Refresh should be disabled on dual plane configs. Otherwise, as the SR watermark is not calculated for such configs, switching to non VGA mode causes FIFO underrun and display flicker.

This fixes Korg Bug #14897.

---
 drivers/gpu/drm/i915/intel_display.c |   12 ++++++++++++
 1 file changed, 12 insertions(+)
drivers/gpu/drm/i915/intel_display.c +12 -0
@@ -2538,6 +2538,10 @@ static void g4x_update_wm(struct drm_dev
 		sr_entries = roundup(sr_entries / cacheline_size, 1);
 		DRM_DEBUG("self-refresh entries: %d\n", sr_entries);
 		I915_WRITE(FW_BLC_SELF, FW_BLC_SELF_EN);
+	} else {
+		/* Turn off self refresh if both pipes are enabled */
+		I915_WRITE(FW_BLC_SELF, I915_READ(FW_BLC_SELF)
+					& ~FW_BLC_SELF_EN);
 	}
 
 	DRM_DEBUG("Setting FIFO watermarks - A: %d, B: %d, SR %d\n",
@@ -2581,6 +2585,10 @@ static void i965_update_wm(struct drm_de
 			srwm = 1;
 		srwm &= 0x3f;
 		I915_WRITE(FW_BLC_SELF, FW_BLC_SELF_EN);
+	} else {
+		/* Turn off self refresh if both pipes are enabled */
+		I915_WRITE(FW_BLC_SELF, I915_READ(FW_BLC_SELF)
+					& ~FW_BLC_SELF_EN);
 	}
 
 	DRM_DEBUG_KMS("Setting FIFO watermarks - A: 8, B: 8, C: 8, SR %d\n",
@@ -2649,6 +2657,10 @@ static void i9xx_update_wm(struct drm_de
 		if (srwm < 0)
 			srwm = 1;
 		I915_WRITE(FW_BLC_SELF, FW_BLC_SELF_EN | (srwm & 0x3f));
+	} else {
+		/* Turn off self refresh if both pipes are enabled */
+		I915_WRITE(FW_BLC_SELF, I915_READ(FW_BLC_SELF)
+					& ~FW_BLC_SELF_EN);
 	}
 
 	DRM_DEBUG("Setting FIFO watermarks - A: %d, B: %d, C: %d, SR %d\n",
author
#13 lore raw

PATCH 13/93 drm/i915: Fix DDC on some systems by clearing BIOS GMBUS setup.

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Eric Anholt <[email protected]>

commit f0217c42c9ab3d772e543f635ce628b9478f70b6 upstream.

This is a sync of a fix I made in the old UMS code. If the BIOS uses the GMBUS and doesn't clear that setup, then our bit-banging I2C can fail, leading to monitors not being detected.

---
 drivers/gpu/drm/i915/i915_reg.h     |   14 ++++++++++++++
 drivers/gpu/drm/i915/i915_suspend.c |    5 ++++-
 drivers/gpu/drm/i915/intel_drv.h    |    2 ++
 drivers/gpu/drm/i915/intel_i2c.c    |   19 +++++++++++++++++++
 4 files changed, 39 insertions(+), 1 deletion(-)
drivers/gpu/drm/i915/i915_reg.h +14 -0
@@ -405,6 +405,13 @@
 # define GPIO_DATA_VAL_IN		(1 << 12)
 # define GPIO_DATA_PULLUP_DISABLE	(1 << 13)
 
+#define GMBUS0			0x5100
+#define GMBUS1			0x5104
+#define GMBUS2			0x5108
+#define GMBUS3			0x510c
+#define GMBUS4			0x5110
+#define GMBUS5			0x5120
+
 /*
  * Clock control & power management
  */
@@ -2153,6 +2160,13 @@
 #define PCH_GPIOE               0xc5020
 #define PCH_GPIOF               0xc5024
 
+#define PCH_GMBUS0		0xc5100
+#define PCH_GMBUS1		0xc5104
+#define PCH_GMBUS2		0xc5108
+#define PCH_GMBUS3		0xc510c
+#define PCH_GMBUS4		0xc5110
+#define PCH_GMBUS5		0xc5120
+
 #define PCH_DPLL_A              0xc6014
 #define PCH_DPLL_B              0xc6018
 
drivers/gpu/drm/i915/i915_suspend.c +4 -1
@@ -27,7 +27,7 @@
 #include "drmP.h"
 #include "drm.h"
 #include "i915_drm.h"
-#include "i915_drv.h"
+#include "intel_drv.h"
 
 static bool i915_pipe_enabled(struct drm_device *dev, enum pipe pipe)
 {
@@ -846,6 +846,9 @@ int i915_restore_state(struct drm_device
 	for (i = 0; i < 3; i++)
 		I915_WRITE(SWF30 + (i << 2), dev_priv->saveSWF2[i]);
 
+	/* I2C state */
+	intel_i2c_reset_gmbus(dev);
+
 	return 0;
 }
 
drivers/gpu/drm/i915/intel_drv.h +2 -0
@@ -134,6 +134,8 @@ void intel_i2c_destroy(struct i2c_adapte
 int intel_ddc_get_modes(struct intel_output *intel_output);
 extern bool intel_ddc_probe(struct intel_output *intel_output);
 void intel_i2c_quirk_set(struct drm_device *dev, bool enable);
+void intel_i2c_reset_gmbus(struct drm_device *dev);
+
 extern void intel_crt_init(struct drm_device *dev);
 extern void intel_hdmi_init(struct drm_device *dev, int sdvox_reg);
 extern bool intel_sdvo_init(struct drm_device *dev, int output_device);
drivers/gpu/drm/i915/intel_i2c.c +19 -0
@@ -118,6 +118,23 @@ static void set_data(void *data, int sta
 	udelay(I2C_RISEFALL_TIME); /* wait for the line to change state */
 }
 
+/* Clears the GMBUS setup.  Our driver doesn't make use of the GMBUS I2C
+ * engine, but if the BIOS leaves it enabled, then that can break our use
+ * of the bit-banging I2C interfaces.  This is notably the case with the
+ * Mac Mini in EFI mode.
+ */
+void
+intel_i2c_reset_gmbus(struct drm_device *dev)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+
+	if (IS_IGDNG(dev)) {
+		I915_WRITE(PCH_GMBUS0, 0);
+	} else {
+		I915_WRITE(GMBUS0, 0);
+	}
+}
+
 /**
  * intel_i2c_create - instantiate an Intel i2c bus using the specified GPIO reg
  * @dev: DRM device
@@ -168,6 +185,8 @@ struct i2c_adapter *intel_i2c_create(str
 	if(i2c_bit_add_bus(&chan->adapter))
 		goto out_free;
 
+	intel_i2c_reset_gmbus(dev);
+
 	/* JJJ:  raise SCL and SDA? */
 	intel_i2c_quirk_set(dev, true);
 	set_data(chan, 1);
author
#14 lore raw

PATCH 14/93 drm/i915: Add HP nx9020/SamsungSX20S to ACPI LID quirk list

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Zhao Yakui <[email protected]>

commit 40f33a92100f4d9b6e85ad642100cfe42d7ff57d upstream.

The HP comaq nx9020/Samsung SX20S laptop always report that the LID status is closed and we can't use it reliabily for LVDS detection. So add the two boxes into the quirk list.

http://bugzilla.kernel.org/show_bug.cgi?id=14957
http://bugzilla.kernel.org/show_bug.cgi?id=14554

---
 drivers/gpu/drm/i915/intel_lvds.c |   14 ++++++++++++++
 1 file changed, 14 insertions(+)
drivers/gpu/drm/i915/intel_lvds.c +14 -0
@@ -602,6 +602,20 @@ static void intel_lvds_mode_set(struct d
 /* Some lid devices report incorrect lid status, assume they're connected */
 static const struct dmi_system_id bad_lid_status[] = {
 	{
+		.ident = "Compaq nx9020",
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "Hewlett-Packard"),
+			DMI_MATCH(DMI_BOARD_NAME, "3084"),
+		},
+	},
+	{
+		.ident = "Samsung SX20S",
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "Phoenix Technologies LTD"),
+			DMI_MATCH(DMI_BOARD_NAME, "SX20S"),
+		},
+	},
+	{
 		.ident = "Aspire One",
 		.matches = {
 			DMI_MATCH(DMI_SYS_VENDOR, "Acer"),
author
#15 lore raw

PATCH 15/93 drm/i915: Fix the incorrect DMI string for Samsung SX20S laptop

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Zhao Yakui <[email protected]>

commit f034b12dbb5749b11e9390e15e93ffa87ece8038 upstream.

---
 drivers/gpu/drm/i915/intel_lvds.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/i915/intel_lvds.c +1 -1
@@ -611,7 +611,7 @@ static const struct dmi_system_id bad_li
 	{
 		.ident = "Samsung SX20S",
 		.matches = {
-			DMI_MATCH(DMI_SYS_VENDOR, "Phoenix Technologies LTD"),
+			DMI_MATCH(DMI_SYS_VENDOR, "Samsung Electronics"),
 			DMI_MATCH(DMI_BOARD_NAME, "SX20S"),
 		},
 	},
author
#16 lore raw

PATCH 16/93 drm/i915: Add MALATA PC-81005 to ACPI LID quirk list

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Zhao Yakui <[email protected]>

commit a3cb5195f6db58dbebd8a31b877ddce082c9b63d upstream.

The MALATA PC-81005 laptop always reports that the LID status is closed and we can't use it reliabily for LVDS detection. So add this box into the quirk list.

https://bugs.freedesktop.org/show_bug.cgi?id=25523

---
 drivers/gpu/drm/i915/intel_lvds.c |    7 +++++++
 1 file changed, 7 insertions(+)
drivers/gpu/drm/i915/intel_lvds.c +7 -0
@@ -622,6 +622,13 @@ static const struct dmi_system_id bad_li
 			DMI_MATCH(DMI_PRODUCT_NAME, "Aspire one"),
 		},
 	},
+	{
+		.ident = "PC-81005",
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "MALATA"),
+			DMI_MATCH(DMI_PRODUCT_NAME, "PC-81005"),
+		},
+	},
 	{ }
 };
 
author
#17 lore raw

PATCH 17/93 usb: r8a66597-hcd: Flush the D-cache for the pipe-in transfer buffers.

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Paul Mundt <[email protected]>

commit 2717568e7c44fe7dc3f4f52ea823811cfeede2b5 upstream.

This implements the same D-cache flushing logic for r8a66597-hcd as Catalin's isp1760 (http://patchwork.kernel.org/patch/76391/) change, with the same note applying here as well:

When the HDC driver writes the data to the transfer buffers it pollutes the D-cache (unlike DMA drivers where the device writes the data). If the corresponding pages get mapped into user space, there are no additional cache flushing operations performed and this causes random user space faults on architectures with separate I and D caches (Harvard) or those with aliasing D-cache.

This fixes up crashes during USB boot on SH7724 and others:

	http://marc.info/?l=linux-sh&m=126439837308912&w=2
---
 drivers/usb/host/r8a66597-hcd.c |   39 ++++++++++++++++++++++++++-------------
 1 file changed, 26 insertions(+), 13 deletions(-)
drivers/usb/host/r8a66597-hcd.c +26 -13
@@ -35,7 +35,9 @@
 #include <linux/usb.h>
 #include <linux/platform_device.h>
 #include <linux/io.h>
+#include <linux/mm.h>
 #include <linux/irq.h>
+#include <asm/cacheflush.h>
 
 #include "../core/hcd.h"
 #include "r8a66597.h"
@@ -820,6 +822,26 @@ static void enable_r8a66597_pipe(struct
 	enable_r8a66597_pipe_dma(r8a66597, dev, pipe, urb);
 }
 
+static void r8a66597_urb_done(struct r8a66597 *r8a66597, struct urb *urb,
+			      int status)
+__releases(r8a66597->lock)
+__acquires(r8a66597->lock)
+{
+	if (usb_pipein(urb->pipe) && usb_pipetype(urb->pipe) != PIPE_CONTROL) {
+		void *ptr;
+
+		for (ptr = urb->transfer_buffer;
+		     ptr < urb->transfer_buffer + urb->transfer_buffer_length;
+		     ptr += PAGE_SIZE)
+			flush_dcache_page(virt_to_page(ptr));
+	}
+
+	usb_hcd_unlink_urb_from_ep(r8a66597_to_hcd(r8a66597), urb);
+	spin_unlock(&r8a66597->lock);
+	usb_hcd_giveback_urb(r8a66597_to_hcd(r8a66597), urb, status);
+	spin_lock(&r8a66597->lock);
+}
+
 /* this function must be called with interrupt disabled */
 static void force_dequeue(struct r8a66597 *r8a66597, u16 pipenum, u16 address)
 {
@@ -840,15 +862,9 @@ static void force_dequeue(struct r8a6659
 		list_del(&td->queue);
 		kfree(td);
 
-		if (urb) {
-			usb_hcd_unlink_urb_from_ep(r8a66597_to_hcd(r8a66597),
-					urb);
-
-			spin_unlock(&r8a66597->lock);
-			usb_hcd_giveback_urb(r8a66597_to_hcd(r8a66597), urb,
-					-ENODEV);
-			spin_lock(&r8a66597->lock);
-		}
+		if (urb)
+			r8a66597_urb_done(r8a66597, urb, -ENODEV);
+
 		break;
 	}
 }
@@ -1285,10 +1301,7 @@ __releases(r8a66597->lock) __acquires(r8
 		if (usb_pipeisoc(urb->pipe))
 			urb->start_frame = r8a66597_get_frame(hcd);
 
-		usb_hcd_unlink_urb_from_ep(r8a66597_to_hcd(r8a66597), urb);
-		spin_unlock(&r8a66597->lock);
-		usb_hcd_giveback_urb(hcd, urb, status);
-		spin_lock(&r8a66597->lock);
+		r8a66597_urb_done(r8a66597, urb, status);
 	}
 
 	if (restart) {
author
#18 lore raw

PATCH 18/93 i2c-tiny-usb: Fix on big-endian systems

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Jean Delvare <[email protected]>

commit 1c010ff8912cbc08d80e865aab9c32b6b00c527d upstream.

The functionality bit vector is always returned as a little-endian 32-bit number by the device, so it must be byte-swapped to the host endianness.

On the other hand, the delay value is handled by the USB stack, so no byte swapping is needed on our side.

This fixes bug #15105:
http://bugzilla.kernel.org/show_bug.cgi?id=15105

---
 drivers/i2c/busses/i2c-tiny-usb.c |   12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)
drivers/i2c/busses/i2c-tiny-usb.c +6 -6
@@ -13,6 +13,7 @@
 #include <linux/kernel.h>
 #include <linux/errno.h>
 #include <linux/module.h>
+#include <linux/types.h>
 
 /* include interfaces to usb layer */
 #include <linux/usb.h>
@@ -31,8 +32,8 @@
 #define CMD_I2C_IO_END		(1<<1)
 
 /* i2c bit delay, default is 10us -> 100kHz */
-static int delay = 10;
-module_param(delay, int, 0);
+static unsigned short delay = 10;
+module_param(delay, ushort, 0);
 MODULE_PARM_DESC(delay, "bit delay in microseconds, "
 		 "e.g. 10 for 100kHz (default is 100kHz)");
 
@@ -109,7 +110,7 @@ static int usb_xfer(struct i2c_adapter *
 
 static u32 usb_func(struct i2c_adapter *adapter)
 {
-	u32 func;
+	__le32 func;
 
 	/* get functionality from adapter */
 	if (usb_read(adapter, CMD_GET_FUNC, 0, 0, &func, sizeof(func)) !=
@@ -118,7 +119,7 @@ static u32 usb_func(struct i2c_adapter *
 		return 0;
 	}
 
-	return func;
+	return le32_to_cpu(func);
 }
 
 /* This is the actual algorithm we define */
@@ -216,8 +217,7 @@ static int i2c_tiny_usb_probe(struct usb
 		 "i2c-tiny-usb at bus %03d device %03d",
 		 dev->usb_dev->bus->busnum, dev->usb_dev->devnum);
 
-	if (usb_write(&dev->adapter, CMD_SET_DELAY,
-		      cpu_to_le16(delay), 0, NULL, 0) != 0) {
+	if (usb_write(&dev->adapter, CMD_SET_DELAY, delay, 0, NULL, 0) != 0) {
 		dev_err(&dev->adapter.dev,
 			"failure setting delay to %dus\n", delay);
 		retval = -EIO;
author
#19 lore raw

PATCH 19/93 drm/i915: handle FBC and self-refresh better

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Jesse Barnes <[email protected]>

commit ee25df2bc379728c45d81e04cf87984db1425edf upstream.

On 945, we need to avoid entering self-refresh if the compressor is busy, or we may cause display FIFO underruns leading to ugly flicker.

Fixes fdo bug #24314, kernel bug #15043.

---
 drivers/gpu/drm/i915/i915_reg.h      |    1 +
 drivers/gpu/drm/i915/intel_display.c |    2 ++
 2 files changed, 3 insertions(+)
drivers/gpu/drm/i915/i915_reg.h +1 -0
@@ -329,6 +329,7 @@
 #define   FBC_CTL_PERIODIC	(1<<30)
 #define   FBC_CTL_INTERVAL_SHIFT (16)
 #define   FBC_CTL_UNCOMPRESSIBLE (1<<14)
+#define   FBC_C3_IDLE		(1<<13)
 #define   FBC_CTL_STRIDE_SHIFT	(5)
 #define   FBC_CTL_FENCENO	(1<<0)
 #define FBC_COMMAND		0x0320c
drivers/gpu/drm/i915/intel_display.c +2 -0
@@ -988,6 +988,8 @@ static void i8xx_enable_fbc(struct drm_c
 
 	/* enable it... */
 	fbc_ctl = FBC_CTL_EN | FBC_CTL_PERIODIC;
+	if (IS_I945GM(dev))
+		fbc_ctl |= FBC_C3_IDLE; /* 945 needs special SR handling */
 	fbc_ctl |= (dev_priv->cfb_pitch & 0xff) << FBC_CTL_STRIDE_SHIFT;
 	fbc_ctl |= (interval & 0x2fff) << FBC_CTL_INTERVAL_SHIFT;
 	if (obj_priv->tiling_mode != I915_TILING_NONE)
author
#20 lore raw

PATCH 20/93 drm/i915: Increase fb alignment to 64k

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Chris Wilson <[email protected]>

commit fd2e8ea597222b8f38ae8948776a61ea7958232e upstream.

An untiled framebuffer must be aligned to 64k. This is normally handled by intel_pin_and_fence_fb_obj(), but the intelfb_create() likes to be different and do the pinning itself. However, it aligns the buffer object incorrectly for pre-i965 chipsets causing a PGTBL_ERR when it is installed onto the output.

Fixes:
KMS error message while initializing modesetting - render error detected: EIR: 0x10 [i915]
http://bugs.freedesktop.org/show_bug.cgi?id=22936

---
 drivers/gpu/drm/i915/intel_fb.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/i915/intel_fb.c +1 -1
@@ -148,7 +148,7 @@ static int intelfb_create(struct drm_dev
 
 	mutex_lock(&dev->struct_mutex);
 
-	ret = i915_gem_object_pin(fbo, PAGE_SIZE);
+	ret = i915_gem_object_pin(fbo, 64*1024);
 	if (ret) {
 		DRM_ERROR("failed to pin fb: %d\n", ret);
 		goto out_unref;
author
#21 lore raw

PATCH 21/93 drm/i915: Update write_domains on active list after flush.

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Daniel Vetter <[email protected]>

commit 99fcb766a3a50466fe31d743260a3400c1aee855 upstream.

Before changing the status of a buffer with a pending write we will await upon a new flush for that buffer. So we can take advantage of any flushes posted whilst the buffer is active and pending processing by the GPU, by clearing its write_domain and updating its last_rendering_seqno -- thus saving a potential flush in deep queues and improves flushing behaviour upon eviction for both GTT space and fences.

In order to reduce the time spent searching the active list for matching write_domains, we move those to a separate list whose elements are the buffers belong to the active/flushing list with pending writes.

Orignal patch by Chris Wilson <[email protected]>, forward-ported by me.

In addition to better performance, this also fixes a real bug. Before this changes, i915_gem_evict_everything didn't work as advertised. When the gpu was actually busy and processing request, the flush and subsequent wait would not move active and dirty buffers to the inactive list, but just to the flushing list. Which triggered the BUG_ON at the end of this function. With the more tight dirty buffer tracking, all currently busy and dirty buffers get moved to the inactive list by one i915_gem_flush operation.

I've left the BUG_ON I've used to prove this in there.

Bug 26101 - [i915] xf86-video-intel 2.10.0 (and git) triggers kernel oops
within seconds after login
http://bugs.freedesktop.org/show_bug.cgi?id=26101

---
 drivers/gpu/drm/i915/i915_drv.h |   11 +++++++++++
 drivers/gpu/drm/i915/i915_gem.c |   23 +++++++++++++++++++----
 2 files changed, 30 insertions(+), 4 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +11 -0
@@ -467,6 +467,15 @@ typedef struct drm_i915_private {
 		struct list_head flushing_list;
 
 		/**
+		 * List of objects currently pending a GPU write flush.
+		 *
+		 * All elements on this list will belong to either the
+		 * active_list or flushing_list, last_rendering_seqno can
+		 * be used to differentiate between the two elements.
+		 */
+		struct list_head gpu_write_list;
+
+		/**
 		 * LRU list of objects which are not in the ringbuffer and
 		 * are ready to unbind, but are still in the GTT.
 		 *
@@ -558,6 +567,8 @@ struct drm_i915_gem_object {
 
 	/** This object's place on the active/flushing/inactive lists */
 	struct list_head list;
+	/** This object's place on GPU write list */
+	struct list_head gpu_write_list;
 
 	/** This object's place on the fenced object LRU */
 	struct list_head fence_list;
drivers/gpu/drm/i915/i915_gem.c +19 -4
@@ -1552,6 +1552,8 @@ i915_gem_object_move_to_inactive(struct
 	else
 		list_move_tail(&obj_priv->list, &dev_priv->mm.inactive_list);
 
+	BUG_ON(!list_empty(&obj_priv->gpu_write_list));
+
 	obj_priv->last_rendering_seqno = 0;
 	if (obj_priv->active) {
 		obj_priv->active = 0;
@@ -1622,7 +1624,8 @@ i915_add_request(struct drm_device *dev,
 		struct drm_i915_gem_object *obj_priv, *next;
 
 		list_for_each_entry_safe(obj_priv, next,
-					 &dev_priv->mm.flushing_list, list) {
+					 &dev_priv->mm.gpu_write_list,
+					 gpu_write_list) {
 			struct drm_gem_object *obj = obj_priv->obj;
 
 			if ((obj->write_domain & flush_domains) ==
@@ -1630,6 +1633,7 @@ i915_add_request(struct drm_device *dev,
 				uint32_t old_write_domain = obj->write_domain;
 
 				obj->write_domain = 0;
+				list_del_init(&obj_priv->gpu_write_list);
 				i915_gem_object_move_to_active(obj, seqno);
 
 				trace_i915_gem_object_change_domain(obj,
@@ -2073,8 +2077,8 @@ static int
 i915_gem_evict_everything(struct drm_device *dev)
 {
 	drm_i915_private_t *dev_priv = dev->dev_private;
-	uint32_t seqno;
 	int ret;
+	uint32_t seqno;
 	bool lists_empty;
 
 	spin_lock(&dev_priv->mm.active_list_lock);
@@ -2096,6 +2100,8 @@ i915_gem_evict_everything(struct drm_dev
 	if (ret)
 		return ret;
 
+	BUG_ON(!list_empty(&dev_priv->mm.flushing_list));
+
 	ret = i915_gem_evict_from_inactive_list(dev);
 	if (ret)
 		return ret;
@@ -2690,7 +2696,7 @@ i915_gem_object_flush_gpu_write_domain(s
 	old_write_domain = obj->write_domain;
 	i915_gem_flush(dev, 0, obj->write_domain);
 	seqno = i915_add_request(dev, NULL, obj->write_domain);
-	obj->write_domain = 0;
+	BUG_ON(obj->write_domain);
 	i915_gem_object_move_to_active(obj, seqno);
 
 	trace_i915_gem_object_change_domain(obj,
@@ -3710,16 +3716,23 @@ i915_gem_execbuffer(struct drm_device *d
 		i915_gem_flush(dev,
 			       dev->invalidate_domains,
 			       dev->flush_domains);
-		if (dev->flush_domains)
+		if (dev->flush_domains & I915_GEM_GPU_DOMAINS)
 			(void)i915_add_request(dev, file_priv,
 					       dev->flush_domains);
 	}
 
 	for (i = 0; i < args->buffer_count; i++) {
 		struct drm_gem_object *obj = object_list[i];
+		struct drm_i915_gem_object *obj_priv = obj->driver_private;
 		uint32_t old_write_domain = obj->write_domain;
 
 		obj->write_domain = obj->pending_write_domain;
+		if (obj->write_domain)
+			list_move_tail(&obj_priv->gpu_write_list,
+				       &dev_priv->mm.gpu_write_list);
+		else
+			list_del_init(&obj_priv->gpu_write_list);
+
 		trace_i915_gem_object_change_domain(obj,
 						    obj->read_domains,
 						    old_write_domain);
@@ -4112,6 +4125,7 @@ int i915_gem_init_object(struct drm_gem_
 	obj_priv->obj = obj;
 	obj_priv->fence_reg = I915_FENCE_REG_NONE;
 	INIT_LIST_HEAD(&obj_priv->list);
+	INIT_LIST_HEAD(&obj_priv->gpu_write_list);
 	INIT_LIST_HEAD(&obj_priv->fence_list);
 	obj_priv->madv = I915_MADV_WILLNEED;
 
@@ -4563,6 +4577,7 @@ i915_gem_load(struct drm_device *dev)
 	spin_lock_init(&dev_priv->mm.active_list_lock);
 	INIT_LIST_HEAD(&dev_priv->mm.active_list);
 	INIT_LIST_HEAD(&dev_priv->mm.flushing_list);
+	INIT_LIST_HEAD(&dev_priv->mm.gpu_write_list);
 	INIT_LIST_HEAD(&dev_priv->mm.inactive_list);
 	INIT_LIST_HEAD(&dev_priv->mm.request_list);
 	INIT_LIST_HEAD(&dev_priv->mm.fence_list);
author
#22 lore raw

PATCH 22/93 regulator: Fix display of null constraints for regulators

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Mark Brown <[email protected]>

commit 973e9a2795b3b41d8408a0bb6f87b783c5efc88a upstream.

If the regulator constraints are empty and there is no voltage reported then nothing will be added to the text displayed for the constraints, leading to random stack data being printed. This is unlikely to happen for practical regulators since most will at least report a voltage but should still be fixed.

---
 drivers/regulator/core.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/regulator/core.c +1 -1
@@ -640,7 +640,7 @@ static int suspend_prepare(struct regula
 static void print_constraints(struct regulator_dev *rdev)
 {
 	struct regulation_constraints *constraints = rdev->constraints;
-	char buf[80];
+	char buf[80] = "";
 	int count;
 
 	if (rdev->desc->type == REGULATOR_VOLTAGE) {
author
#23 lore raw

PATCH 23/93 ALSA: hda-intel: Avoid divide by zero crash

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Jody Bruchon <[email protected]>

commit fed08d036f2aabd8d0c684439de37f8ebec2bbc2 upstream.

On my AMD780V chipset, hda_intel.c can crash the kernel with a divide by zero
for as-yet unknown reasons. A simple check for zero prevents it, though the problem that causes it remains. Since the workaround is harmless and won't affect anyone except victims of this bug, it should be safe; moreover,
because this crash can be triggered by a user-mode application, there are denial of service implications on the systems affected by the bug without the patch.

---
 sound/pci/hda/hda_intel.c |    6 ++++++
 1 file changed, 6 insertions(+)
sound/pci/hda/hda_intel.c +6 -0
@@ -1858,6 +1858,12 @@ static int azx_position_ok(struct azx *c
 
 	if (!bdl_pos_adj[chip->dev_index])
 		return 1; /* no delayed ack */
+	if (azx_dev->period_bytes == 0) {
+		printk(KERN_WARNING
+		       "hda-intel: Divide by zero was avoided "
+		       "in azx_dev->period_bytes.\n");
+		return 0;
+	}
 	if (pos % azx_dev->period_bytes > azx_dev->period_bytes / 2)
 		return 0; /* NG - it's below the period boundary */
 	return 1; /* OK, it's fine */
author
#24 lore raw

PATCH 24/93 CPUFREQ: Fix use after free of struct powernow_k8_data

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Thomas Renninger <[email protected]>

commit 557a701c16553b0b691dbb64ef30361115a80f64 upstream.

Easy fix for a regression introduced in 2.6.31.

On managed CPUs the cpufreq.c core will call driver->exit(cpu) on the managed cpus and powernow_k8 will free the core's data.

Later driver->get(cpu) function might get called trying to read out the current freq of a managed cpu and the NULL pointer check does not work on the freed object -> better set it to NULL.

->get() is unsigned and must return 0 as invalid frequency.

Reference:
http://bugzilla.kernel.org/show_bug.cgi?id=14391

---
 arch/x86/kernel/cpu/cpufreq/powernow-k8.c |    3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
arch/x86/kernel/cpu/cpufreq/powernow-k8.c +2 -1
@@ -1351,6 +1351,7 @@ static int __devexit powernowk8_cpu_exit
 
 	kfree(data->powernow_table);
 	kfree(data);
+	per_cpu(powernow_data, pol->cpu) = NULL;
 
 	return 0;
 }
@@ -1370,7 +1371,7 @@ static unsigned int powernowk8_get(unsig
 	int err;
 
 	if (!data)
-		return -EINVAL;
+		return 0;
 
 	smp_call_function_single(cpu, query_values_on_cpu, &err, true);
 	if (err)
author
#25 lore raw

PATCH 25/93 freeze_bdev: dont deactivate successfully frozen MS_RDONLY sb

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Jun'ichi Nomura <[email protected]>

commit 4b06e5b9ad8abb20105b2b25e42c509ebe9b2d76 upstream.

Thanks Thomas and Christoph for testing and review.
I removed 'smp_wmb()' before up_write from the previous patch, since up_write() should have necessary ordering constraints. (I.e. the change of s_frozen is visible to others after up_write) I'm quite sure the change is harmless but if you are uncomfortable with Tested-by/Reviewed-by on the modified patch, please remove them.

If MS_RDONLY, freeze_bdev should just up_write(s_umount) instead of deactivate_locked_super().
Also, keep sb->s_frozen consistent so that remount can check the frozen state.

Otherwise a crash reported here can happen:
http://lkml.org/lkml/2010/1/16/37
http://lkml.org/lkml/2010/1/28/53

This patch should be applied for 2.6.32 stable series, too.

fs/block_dev.c +4 -3
@@ -246,7 +246,8 @@ struct super_block *freeze_bdev(struct block_device *bdev)
 	if (!sb)
 		goto out;
 	if (sb->s_flags & MS_RDONLY) {
-		deactivate_locked_super(sb);
+		sb->s_frozen = SB_FREEZE_TRANS;
+		up_write(&sb->s_umount);
 		mutex_unlock(&bdev->bd_fsfreeze_mutex);
 		return sb;
 	}
@@ -307,7 +308,7 @@ int thaw_bdev(struct block_device *bdev, struct super_block *sb)
 	BUG_ON(sb->s_bdev != bdev);
 	down_write(&sb->s_umount);
 	if (sb->s_flags & MS_RDONLY)
-		goto out_deactivate;
+		goto out_unfrozen;
 
 	if (sb->s_op->unfreeze_fs) {
 		error = sb->s_op->unfreeze_fs(sb);
@@ -321,11 +322,11 @@ int thaw_bdev(struct block_device *bdev, struct super_block *sb)
 		}
 	}
 
+out_unfrozen:
 	sb->s_frozen = SB_UNFROZEN;
 	smp_wmb();
 	wake_up(&sb->s_wait_unfrozen);
 
-out_deactivate:
 	if (sb)
 		deactivate_locked_super(sb);
 out_unlock:
author
#26 lore raw

PATCH 26/93 cciss: Make cciss_seq_show handle holes in the h->drv[] array

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Stephen M. Cameron <[email protected]>

commit 531c2dc70d339c5dfa8c3eb628c3459dc6f3a075 upstream.

It is possible (and expected) for there to be holes in the h->drv[] array, that is, some elements may be NULL pointers. cciss_seq_show needs to be made aware of this possibility to avoid an Oops.

To reproduce the Oops which this fixes:

1) Create two "arrays" in the Array Configuratino Utility and several logical drives on each array.
2) cat /proc/driver/cciss/cciss* in an infinite loop
3) delete some of the logical drives in the first "array."

---
 drivers/block/cciss.c |    3 +++
 1 file changed, 3 insertions(+)
drivers/block/cciss.c +3 -0
@@ -339,6 +339,9 @@ static int cciss_seq_show(struct seq_fil
 	if (*pos > h->highest_lun)
 		return 0;
 
+	if (drv == NULL) /* it's possible for h->drv[] to have holes. */
+		return 0;
+
 	if (drv->heads == 0)
 		return 0;
 
author
#27 lore raw

PATCH 27/93 ioat: fix infinite timeout checking in ioat2_quiesce

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Dan Williams <[email protected]>

commit 7e55a70c5b9a57c12f49c44b0847c9343d4f54e4 upstream.

Fix typo in ioat2_quiesce. check 'tmo' is zero, not 'end'. Also applies to 2.6.32.3

---
 drivers/dma/ioat/dma_v2.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/dma/ioat/dma_v2.c +1 -1
@@ -249,7 +249,7 @@ int ioat2_quiesce(struct ioat_chan_commo
 	if (is_ioat_active(status) || is_ioat_idle(status))
 		ioat_suspend(chan);
 	while (is_ioat_active(status) || is_ioat_idle(status)) {
-		if (end && time_after(jiffies, end)) {
+		if (tmo && time_after(jiffies, end)) {
 			err = -ETIMEDOUT;
 			break;
 		}
author
#28 lore raw

PATCH 28/93 resource: add helpers for fetching rlimits

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Jiri Slaby <[email protected]>

commit 3e10e716abf3c71bdb5d86b8f507f9e72236c9cd upstream.

We want to be sure that compiler fetches the limit variable only once, so add helpers for fetching current and maximal resource limits which do that.

Add them to sched.h (instead of resource.h) due to circular dependency
sched.h->resource.h->task_struct
Alternative would be to create a separate res_access.h or similar.

Cc James Morris <[email protected]>
Cc Heiko Carstens <[email protected]>
Cc Andrew Morton <[email protected]>
Cc Ingo Molnar <[email protected]>
---
 include/linux/sched.h |   22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)
include/linux/sched.h +22 -0
@@ -2583,6 +2583,28 @@ static inline void mm_init_owner(struct
 
 #define TASK_STATE_TO_CHAR_STR "RSDTtZX"
 
+static inline unsigned long task_rlimit(const struct task_struct *tsk,
+		unsigned int limit)
+{
+	return ACCESS_ONCE(tsk->signal->rlim[limit].rlim_cur);
+}
+
+static inline unsigned long task_rlimit_max(const struct task_struct *tsk,
+		unsigned int limit)
+{
+	return ACCESS_ONCE(tsk->signal->rlim[limit].rlim_max);
+}
+
+static inline unsigned long rlimit(unsigned int limit)
+{
+	return task_rlimit(current, limit);
+}
+
+static inline unsigned long rlimit_max(unsigned int limit)
+{
+	return task_rlimit_max(current, limit);
+}
+
 #endif /* __KERNEL__ */
 
 #endif
author
#29 lore raw

PATCH 29/93 fs/exec.c: restrict initial stack space expansion to rlimit

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Michael Neuling <[email protected]>

commit 803bf5ec259941936262d10ecc84511b76a20921 upstream.

When reserving stack space for a new process, make sure we're not attempting to expand the stack by more than rlimit allows.

This fixes a bug caused by b6a2fea39318e43fee84fa7b0b90d68bed92d2ba ("mm:
variable length argument support") and unmasked by fc63cf237078c86214abcb2ee9926d8ad289da9b ("exec: setup_arg_pages() fails to return errors").

This bug means that when limiting the stack to less the 20*PAGE_SIZE (eg. 80K on 4K pages or 'ulimit -s 79') all processes will be killed before they start. This is particularly bad with 64K pages, where a ulimit below 1280K will kill every process.

To test, do:

'ulimit -s 15; ls'

before and after the patch is applied. Before it's applied, 'ls' should be killed. After the patch is applied, 'ls' should no longer be killed.

A stack limit of 15KB since it's small enough to trigger 20*PAGE_SIZE. Also 15KB not a multiple of PAGE_SIZE, which is a trickier case to handle correctly with this code.

4K pages should be fine to test with.

[[email protected]: cleanup cleanup]
Cc Americo Wang <[email protected]>
Cc Anton Blanchard <[email protected]>
Cc Oleg Nesterov <[email protected]>
Cc James Morris <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Serge Hallyn <[email protected]>
Cc Benjamin Herrenschmidt <[email protected]>
---
 fs/exec.c |   21 +++++++++++++++++++--
 1 file changed, 19 insertions(+), 2 deletions(-)
fs/exec.c +19 -2
@@ -572,6 +572,9 @@ int setup_arg_pages(struct linux_binprm
 	struct vm_area_struct *prev = NULL;
 	unsigned long vm_flags;
 	unsigned long stack_base;
+	unsigned long stack_size;
+	unsigned long stack_expand;
+	unsigned long rlim_stack;
 
 #ifdef CONFIG_STACK_GROWSUP
 	/* Limit stack size to 1GB */
@@ -628,10 +631,24 @@ int setup_arg_pages(struct linux_binprm
 			goto out_unlock;
 	}
 
+	stack_expand = EXTRA_STACK_VM_PAGES * PAGE_SIZE;
+	stack_size = vma->vm_end - vma->vm_start;
+	/*
+	 * Align this down to a page boundary as expand_stack
+	 * will align it up.
+	 */
+	rlim_stack = rlimit(RLIMIT_STACK) & PAGE_MASK;
+	rlim_stack = min(rlim_stack, stack_size);
 #ifdef CONFIG_STACK_GROWSUP
-	stack_base = vma->vm_end + EXTRA_STACK_VM_PAGES * PAGE_SIZE;
+	if (stack_size + stack_expand > rlim_stack)
+		stack_base = vma->vm_start + rlim_stack;
+	else
+		stack_base = vma->vm_end + stack_expand;
 #else
-	stack_base = vma->vm_start - EXTRA_STACK_VM_PAGES * PAGE_SIZE;
+	if (stack_size + stack_expand > rlim_stack)
+		stack_base = vma->vm_end - rlim_stack;
+	else
+		stack_base = vma->vm_start - stack_expand;
 #endif
 	ret = expand_stack(vma, stack_base);
 	if (ret)
#30 lore raw

Re: [29/93] fs/exec.c: restrict initial stack space expansion to rlimit

Quoted 1 line · In message <[email protected]> you wrote:

2.6.32-stable review patch. If anyone has any objections, please let us know

.

Quoted 9 lines

------------------

From: Michael Neuling <[email protected]>

commit 803bf5ec259941936262d10ecc84511b76a20921 upstream.

When reserving stack space for a new process, make sure we're not attempting to expand the stack by more than rlimit allows.

This breaks UML, so you also need to take this also:

http://patchwork.kernel.org/patch/79365/

It's in akpm's tree only so far.

Mikey

Quoted 84 lines

This fixes a bug caused by b6a2fea39318e43fee84fa7b0b90d68bed92d2ba ("mm:
variable length argument support") and unmasked by fc63cf237078c86214abcb2ee9926d8ad289da9b ("exec: setup_arg_pages() fails to return errors").

This bug means that when limiting the stack to less the 20*PAGE_SIZE (eg. 80K on 4K pages or 'ulimit -s 79') all processes will be killed before they start. This is particularly bad with 64K pages, where a ulimit below 1280K will kill every process.

To test, do:

'ulimit -s 15; ls'

before and after the patch is applied. Before it's applied, 'ls' should be killed. After the patch is applied, 'ls' should no longer be killed.

A stack limit of 15KB since it's small enough to trigger 20*PAGE_SIZE. Also 15KB not a multiple of PAGE_SIZE, which is a trickier case to handle correctly with this code.

4K pages should be fine to test with.

[[email protected]: cleanup cleanup]
Cc Americo Wang <[email protected]>
Cc Anton Blanchard <[email protected]>
Cc Oleg Nesterov <[email protected]>
Cc James Morris <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Serge Hallyn <[email protected]>
Cc Benjamin Herrenschmidt <[email protected]>
---
 fs/exec.c |   21 +++++++++++++++++++--
 1 file changed, 19 insertions(+), 2 deletions(-)
fs/exec.c +19 -2
@@ -572,6 +572,9 @@ int setup_arg_pages(struct linux_binprm
 	struct vm_area_struct *prev = NULL;
 	unsigned long vm_flags;
 	unsigned long stack_base;
+	unsigned long stack_size;
+	unsigned long stack_expand;
+	unsigned long rlim_stack;
 
 #ifdef CONFIG_STACK_GROWSUP
 	/* Limit stack size to 1GB */
@@ -628,10 +631,24 @@ int setup_arg_pages(struct linux_binprm
 			goto out_unlock;
 	}
 
+	stack_expand = EXTRA_STACK_VM_PAGES * PAGE_SIZE;
+	stack_size = vma->vm_end - vma->vm_start;
+	/*
+	 * Align this down to a page boundary as expand_stack
+	 * will align it up.
+	 */
+	rlim_stack = rlimit(RLIMIT_STACK) & PAGE_MASK;
+	rlim_stack = min(rlim_stack, stack_size);
 #ifdef CONFIG_STACK_GROWSUP
-	stack_base = vma->vm_end + EXTRA_STACK_VM_PAGES * PAGE_SIZE;
+	if (stack_size + stack_expand > rlim_stack)
+		stack_base = vma->vm_start + rlim_stack;
+	else
+		stack_base = vma->vm_end + stack_expand;
 #else
-	stack_base = vma->vm_start - EXTRA_STACK_VM_PAGES * PAGE_SIZE;
+	if (stack_size + stack_expand > rlim_stack)
+		stack_base = vma->vm_end - rlim_stack;
+	else
+		stack_base = vma->vm_start - stack_expand;
 #endif
 	ret = expand_stack(vma, stack_base);
 	if (ret)
#31 lore raw

Re: [stable] [29/93] fs/exec.c: restrict initial stack space expansion to rlimit

Quoted 10 lines · Michael Neuling
Quoted 1 line · In message <[email protected]> you wrote:

2.6.32-stable review patch. If anyone has any objections, please let us know

.

Quoted 7 lines

------------------

From: Michael Neuling <[email protected]>

commit 803bf5ec259941936262d10ecc84511b76a20921 upstream.

Quoted 8 lines
Quoted 2 lines

When reserving stack space for a new process, make sure we're not attempting to expand the stack by more than rlimit allows.

This breaks UML, so you also need to take this also:

http://patchwork.kernel.org/patch/79365/

It's in akpm's tree only so far.

As I can't take patches that are not in Linus's tree, can you let [email protected] know the git commit id of the patch when it goes into Linus's tree, and then I will apply it to the next .32-stable release?

thanks,

greg k-h

#32 lore raw

Re: [stable] [29/93] fs/exec.c: restrict initial stack space expansion to rlimit

Quoted 3 lines · In message <[email protected]> you wrote:
Quoted 2 lines · Michael Neuling
Quoted 1 line · In message <[email protected]> you wrote:

2.6.32-stable review patch. If anyone has any objections, please let us

know

Quoted 13 lines
Quoted 13 lines

.

Quoted 9 lines

------------------

From: Michael Neuling <[email protected]>

commit 803bf5ec259941936262d10ecc84511b76a20921 upstream.

When reserving stack space for a new process, make sure we're not attempting to expand the stack by more than rlimit allows.

This breaks UML, so you also need to take this also:

Quoted 7 lines
Quoted 3 lines

http://patchwork.kernel.org/patch/79365/

It's in akpm's tree only so far.

As I can't take patches that are not in Linus's tree, can you let [email protected] know the git commit id of the patch when it goes into Linus's tree, and then I will apply it to the next .32-stable release?

No problems. Actually, Linus took it over night. SHA1 is a17e18790a8c47113a73139d54a375dc9ccd8f08 "fs/exec.c: fix initial stack reservation"

I'll send a note to stable now.

Mikey

author
#33 lore raw

PATCH 30/93 cifs: fix length calculation for converted unicode readdir names

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Jeff Layton <[email protected]>

commit f12f98dba6ea1517cd7fbb912208893b9c014c15 upstream.

cifs_from_ucs2 returns the length of the converted name, including the length of the NULL terminator. We don't want to include the NULL terminator in the dentry name length however since that'll throw off the hash calculation for the dentry cache.

I believe that this is the root cause of several problems that have cropped up recently that seem to be papered over with the "noserverino" mount option. More confirmation of that would be good, but this is clearly a bug and it fixes at least one reproducible problem that was reported.

This patch fixes at least this reproducer in this kernel.org bug:

    http://bugzilla.kernel.org/show_bug.cgi?id=15088#c12
---
 fs/cifs/readdir.c |    1 +
 1 file changed, 1 insertion(+)
fs/cifs/readdir.c +1 -0
@@ -666,6 +666,7 @@ static int cifs_get_name_from_search_buf
 					   min(len, max_len), nlt,
 					   cifs_sb->mnt_cifs_flags &
 						CIFS_MOUNT_MAP_SPECIAL_CHR);
+		pqst->len -= nls_nullsize(nlt);
 	} else {
 		pqst->name = filename;
 		pqst->len = len;
author
#34 lore raw

PATCH 31/93 NFS: Fix a reference leak in nfs_wb_cancel_page()

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Trond Myklebust <[email protected]>

commit c9edda7140ec6a22accf7f2f86da362dfbfd41fc upstream.

---
 fs/nfs/write.c |    1 +
 1 file changed, 1 insertion(+)
fs/nfs/write.c +1 -0
@@ -1542,6 +1542,7 @@ int nfs_wb_page_cancel(struct inode *ino
 			break;
 		}
 		ret = nfs_wait_on_request(req);
+		nfs_release_request(req);
 		if (ret < 0)
 			goto out;
 	}
author
#35 lore raw

PATCH 32/93 NFS: Try to commit unstable writes in nfs_release_page()

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Trond Myklebust <[email protected]>

commit 82be934a59ff891cac598727e5a862ba2b9d1fac upstream.

If someone calls nfs_release_page(), we presumably already know that the page is clean, however it may be holding an unstable write.

---
 fs/nfs/file.c |    2 ++
 1 file changed, 2 insertions(+)
fs/nfs/file.c +2 -0
@@ -486,6 +486,8 @@ static int nfs_release_page(struct page
 {
 	dfprintk(PAGECACHE, "NFS: release_page(%p)\n", page);
 
+	if (gfp & __GFP_WAIT)
+		nfs_wb_page(page->mapping->host, page);
 	/* If PagePrivate() is set, then the page is not freeable */
 	if (PagePrivate(page))
 		return 0;
author
#36 lore raw

PATCH 33/93 NFSv4: Dont allow posix locking against servers that dont support it

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Trond Myklebust <[email protected]>

commit 8e469ebd6dc32cbaf620e134d79f740bf0ebab79 upstream.

---
 fs/nfs/nfs4_fs.h  |    1 +
 fs/nfs/nfs4proc.c |    7 ++++++-
 2 files changed, 7 insertions(+), 1 deletion(-)
fs/nfs/nfs4_fs.h +1 -0
@@ -141,6 +141,7 @@ enum {
 	NFS_O_RDWR_STATE,		/* OPEN stateid has read/write state */
 	NFS_STATE_RECLAIM_REBOOT,	/* OPEN stateid server rebooted */
 	NFS_STATE_RECLAIM_NOGRACE,	/* OPEN stateid needs to recover state */
+	NFS_STATE_POSIX_LOCKS,		/* Posix locks are supported */
 };
 
 struct nfs4_state {
fs/nfs/nfs4proc.c +6 -1
@@ -1573,6 +1573,8 @@ static int _nfs4_do_open(struct inode *d
 	status = PTR_ERR(state);
 	if (IS_ERR(state))
 		goto err_opendata_put;
+	if ((opendata->o_res.rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) != 0)
+		set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
 	nfs4_opendata_put(opendata);
 	nfs4_put_state_owner(sp);
 	*res = state;
@@ -4060,8 +4062,11 @@ static int _nfs4_proc_setlk(struct nfs4_
 {
 	struct nfs_inode *nfsi = NFS_I(state->inode);
 	unsigned char fl_flags = request->fl_flags;
-	int status;
+	int status = -ENOLCK;
 
+	if ((fl_flags & FL_POSIX) &&
+			!test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
+		goto out;
 	/* Is this a delegated open? */
 	status = nfs4_set_lock_state(state, request);
 	if (status != 0)
author
#37 lore raw

PATCH 34/93 NFSv4: Ensure that the NFSv4 locking can recover from stateid errors

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Trond Myklebust <[email protected]>

commit 2bee72a6aa1e6d0a4f5da56217f0d0bbbdd0d9a3 upstream.

In most cases, we just want to mark the lock_stateid sequence id as being uninitialised.

---
 fs/nfs/nfs4proc.c |   19 +++++++++++++++++++
 1 file changed, 19 insertions(+)
fs/nfs/nfs4proc.c +19 -0
@@ -3978,6 +3978,22 @@ static const struct rpc_call_ops nfs4_lo
 	.rpc_release = nfs4_lock_release,
 };
 
+static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
+{
+	struct nfs_client *clp = server->nfs_client;
+	struct nfs4_state *state = lsp->ls_state;
+
+	switch (error) {
+	case -NFS4ERR_ADMIN_REVOKED:
+	case -NFS4ERR_BAD_STATEID:
+	case -NFS4ERR_EXPIRED:
+		if (new_lock_owner != 0 ||
+		   (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
+			nfs4_state_mark_reclaim_nograce(clp, state);
+		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
+	};
+}
+
 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int reclaim)
 {
 	struct nfs4_lockdata *data;
@@ -4013,6 +4029,9 @@ static int _nfs4_do_setlk(struct nfs4_st
 	ret = nfs4_wait_for_completion_rpc_task(task);
 	if (ret == 0) {
 		ret = data->rpc_status;
+		if (ret)
+			nfs4_handle_setlk_error(data->server, data->lsp,
+					data->arg.new_lock_owner, ret);
 	} else
 		data->cancelled = 1;
 	rpc_put_task(task);
author
#38 lore raw

PATCH 35/93 NFS: Fix an Oops when truncating a file

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Trond Myklebust <[email protected]>

commit 9f557cd8073104b39528794d44e129331ded649f upstream.

The VM/VFS does not allow mapping->a_ops->invalidatepage() to fail. Unfortunately, nfs_wb_page_cancel() may fail if a fatal signal occurs. Since the NFS code assumes that the page stays mapped for as long as the writeback is active, we can end up Oopsing (among other things).

The only safe fix here is to convert nfs_wait_on_request(), so as to make it uninterruptible (as is already the case with wait_on_page_writeback()).

fs/nfs/pagelist.c +9 -8
@@ -176,6 +176,12 @@ void nfs_release_request(struct nfs_page *req)
 	kref_put(&req->wb_kref, nfs_free_request);
 }
 
+static int nfs_wait_bit_uninterruptible(void *word)
+{
+	io_schedule();
+	return 0;
+}
+
 /**
  * nfs_wait_on_request - Wait for a request to complete.
  * @req: request to wait upon.
@@ -186,14 +192,9 @@ void nfs_release_request(struct nfs_page *req)
 int
 nfs_wait_on_request(struct nfs_page *req)
 {
-	int ret = 0;
-
-	if (!test_bit(PG_BUSY, &req->wb_flags))
-		goto out;
-	ret = out_of_line_wait_on_bit(&req->wb_flags, PG_BUSY,
-			nfs_wait_bit_killable, TASK_KILLABLE);
-out:
-	return ret;
+	return wait_on_bit(&req->wb_flags, PG_BUSY,
+			nfs_wait_bit_uninterruptible,
+			TASK_UNINTERRUPTIBLE);
 }
 
 /**
author
#39 lore raw

PATCH 36/93 NFS: Fix a umount race

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Trond Myklebust <[email protected]>

commit 387c149b54b4321cbc790dadbd4f8eedb5a90468 upstream.

Ensure that we unregister the bdi before kill_anon_super() calls ida_remove() on our device name.

---
 fs/nfs/super.c |   15 ++++++++++++++-
 1 file changed, 14 insertions(+), 1 deletion(-)
fs/nfs/super.c +14 -1
@@ -241,6 +241,7 @@ static int  nfs_show_stats(struct seq_fi
 static int nfs_get_sb(struct file_system_type *, int, const char *, void *, struct vfsmount *);
 static int nfs_xdev_get_sb(struct file_system_type *fs_type,
 		int flags, const char *dev_name, void *raw_data, struct vfsmount *mnt);
+static void nfs_put_super(struct super_block *);
 static void nfs_kill_super(struct super_block *);
 static int nfs_remount(struct super_block *sb, int *flags, char *raw_data);
 
@@ -264,6 +265,7 @@ static const struct super_operations nfs
 	.alloc_inode	= nfs_alloc_inode,
 	.destroy_inode	= nfs_destroy_inode,
 	.write_inode	= nfs_write_inode,
+	.put_super	= nfs_put_super,
 	.statfs		= nfs_statfs,
 	.clear_inode	= nfs_clear_inode,
 	.umount_begin	= nfs_umount_begin,
@@ -333,6 +335,7 @@ static const struct super_operations nfs
 	.alloc_inode	= nfs_alloc_inode,
 	.destroy_inode	= nfs_destroy_inode,
 	.write_inode	= nfs_write_inode,
+	.put_super	= nfs_put_super,
 	.statfs		= nfs_statfs,
 	.clear_inode	= nfs4_clear_inode,
 	.umount_begin	= nfs_umount_begin,
@@ -2196,6 +2199,17 @@ error_splat_super:
 }
 
 /*
+ * Ensure that we unregister the bdi before kill_anon_super
+ * releases the device name
+ */
+static void nfs_put_super(struct super_block *s)
+{
+	struct nfs_server *server = NFS_SB(s);
+
+	bdi_unregister(&server->backing_dev_info);
+}
+
+/*
  * Destroy an NFS2/3 superblock
  */
 static void nfs_kill_super(struct super_block *s)
@@ -2203,7 +2217,6 @@ static void nfs_kill_super(struct super_
 	struct nfs_server *server = NFS_SB(s);
 
 	kill_anon_super(s);
-	bdi_unregister(&server->backing_dev_info);
 	nfs_fscache_release_super_cookie(s);
 	nfs_free_server(server);
 }
author
#40 lore raw

PATCH 37/93 NFS: Fix a bug in nfs_fscache_release_page()

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Trond Myklebust <[email protected]>

commit 2c1740098c708b465e87637b237feb2fd98f129a upstream.

Not having an fscache cookie is perfectly valid if the user didn't mount with the fscache option.

This patch fixes http://bugzilla.kernel.org/show_bug.cgi?id=15234

---
 fs/nfs/fscache.c |    9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)
fs/nfs/fscache.c +4 -5
@@ -354,12 +354,11 @@ void nfs_fscache_reset_inode_cookie(stru
  */
 int nfs_fscache_release_page(struct page *page, gfp_t gfp)
 {
-	struct nfs_inode *nfsi = NFS_I(page->mapping->host);
-	struct fscache_cookie *cookie = nfsi->fscache;
-
-	BUG_ON(!cookie);
-
 	if (PageFsCache(page)) {
+		struct nfs_inode *nfsi = NFS_I(page->mapping->host);
+		struct fscache_cookie *cookie = nfsi->fscache;
+
+		BUG_ON(!cookie);
 		dfprintk(FSCACHE, "NFS: fscache releasepage (0x%p/0x%p/0x%p)\n",
 			 cookie, page, nfsi);
 
author
#41 lore raw

PATCH 38/93 NFS: Fix the mapping of the NFSERR_SERVERFAULT error

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Trond Myklebust <[email protected]>

commit fdcb45777a3d1689c5541e1f85ee3ebbd197d2c1 upstream.

It was recently pointed out that the NFSERR_SERVERFAULT error, which is designed to inform the user of a serious internal error on the server, was being mapped to an error value that is internal to the kernel.

This patch maps it to the error EREMOTEIO, which is exported to userland through errno.h.

---
 fs/nfs/mount_clnt.c |    2 +-
 fs/nfs/nfs2xdr.c    |    2 +-
 fs/nfs/nfs4xdr.c    |    6 +++---
 3 files changed, 5 insertions(+), 5 deletions(-)
fs/nfs/mount_clnt.c +1 -1
@@ -120,7 +120,7 @@ static struct {
 	{ .status = MNT3ERR_INVAL,		.errno = -EINVAL,	},
 	{ .status = MNT3ERR_NAMETOOLONG,	.errno = -ENAMETOOLONG,	},
 	{ .status = MNT3ERR_NOTSUPP,		.errno = -ENOTSUPP,	},
-	{ .status = MNT3ERR_SERVERFAULT,	.errno = -ESERVERFAULT,	},
+	{ .status = MNT3ERR_SERVERFAULT,	.errno = -EREMOTEIO,	},
 };
 
 struct mountres {
fs/nfs/nfs2xdr.c +1 -1
@@ -699,7 +699,7 @@ static struct {
 	{ NFSERR_BAD_COOKIE,	-EBADCOOKIE	},
 	{ NFSERR_NOTSUPP,	-ENOTSUPP	},
 	{ NFSERR_TOOSMALL,	-ETOOSMALL	},
-	{ NFSERR_SERVERFAULT,	-ESERVERFAULT	},
+	{ NFSERR_SERVERFAULT,	-EREMOTEIO	},
 	{ NFSERR_BADTYPE,	-EBADTYPE	},
 	{ NFSERR_JUKEBOX,	-EJUKEBOX	},
 	{ -1,			-EIO		}
fs/nfs/nfs4xdr.c +3 -3
@@ -4554,7 +4554,7 @@ static int decode_sequence(struct xdr_st
 	 * If the server returns different values for sessionID, slotID or
 	 * sequence number, the server is looney tunes.
 	 */
-	status = -ESERVERFAULT;
+	status = -EREMOTEIO;
 
 	if (memcmp(id.data, res->sr_session->sess_id.data,
 		   NFS4_MAX_SESSIONID_LEN)) {
@@ -5678,7 +5678,7 @@ static struct {
 	{ NFS4ERR_BAD_COOKIE,	-EBADCOOKIE	},
 	{ NFS4ERR_NOTSUPP,	-ENOTSUPP	},
 	{ NFS4ERR_TOOSMALL,	-ETOOSMALL	},
-	{ NFS4ERR_SERVERFAULT,	-ESERVERFAULT	},
+	{ NFS4ERR_SERVERFAULT,	-EREMOTEIO	},
 	{ NFS4ERR_BADTYPE,	-EBADTYPE	},
 	{ NFS4ERR_LOCKED,	-EAGAIN		},
 	{ NFS4ERR_SYMLINK,	-ELOOP		},
@@ -5705,7 +5705,7 @@ nfs4_stat_to_errno(int stat)
 	}
 	if (stat <= 10000 || stat > 10100) {
 		/* The server is looney tunes. */
-		return -ESERVERFAULT;
+		return -EREMOTEIO;
 	}
 	/* If we cannot translate the error, the recovery routines should
 	 * handle it.
author
#42 lore raw

PATCH 39/93 md: fix degraded calculation when starting a reshape.

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: NeilBrown <[email protected]>

commit 9eb07c259207d048e3ee8be2a77b2a4680b1edd4 upstream.

This code was written long ago when it was not possible to reshape a degraded array. Now it is so the current level of degraded-ness needs to be taken in to account. Also newly addded devices should only reduce degradedness if they are deemed to be in-sync.

In particular, if you convert a RAID5 to a RAID6, and increase the number of devices at the same time, then the 5->6 conversion will make the array degraded so the current code will produce a wrong value for 'degraded' - "-1" to be precise.

If the reshape runs to completion end_reshape will calculate a correct new value for 'degraded', but if a device fails during the reshape an incorrect decision might be made based on the incorrect value of "degraded".

This patch is suitable for 2.6.32-stable and if they are still open, 2.6.31-stable and 2.6.30-stable as well.

---
 drivers/md/raid5.c |   11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)
drivers/md/raid5.c +7 -4
@@ -5432,11 +5432,11 @@ static int raid5_start_reshape(mddev_t *
 		    !test_bit(Faulty, &rdev->flags)) {
 			if (raid5_add_disk(mddev, rdev) == 0) {
 				char nm[20];
-				if (rdev->raid_disk >= conf->previous_raid_disks)
+				if (rdev->raid_disk >= conf->previous_raid_disks) {
 					set_bit(In_sync, &rdev->flags);
-				else
+					added_devices++;
+				} else
 					rdev->recovery_offset = 0;
-				added_devices++;
 				sprintf(nm, "rd%d", rdev->raid_disk);
 				if (sysfs_create_link(&mddev->kobj,
 						      &rdev->kobj, nm))
@@ -5448,9 +5448,12 @@ static int raid5_start_reshape(mddev_t *
 				break;
 		}
 
+	/* When a reshape changes the number of devices, ->degraded
+	 * is measured against the large of the pre and post number of
+	 * devices.*/
 	if (mddev->delta_disks > 0) {
 		spin_lock_irqsave(&conf->device_lock, flags);
-		mddev->degraded = (conf->raid_disks - conf->previous_raid_disks)
+		mddev->degraded += (conf->raid_disks - conf->previous_raid_disks)
 			- added_devices;
 		spin_unlock_irqrestore(&conf->device_lock, flags);
 	}
author
#43 lore raw

PATCH 40/93 V4L/DVB: dvb-core: fix initialization of feeds list in demux filter

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Francesco Lavra <[email protected]>

commit 691c9ae099b9bcb5c27125af00a4a90120977458 upstream.

A DVB demultiplexer device can be used to set up either a PES filter or a section filter. In the former case, the ts field of the feed union of struct dmxdev_filter is used, in the latter case the sec field of the same union is used.
The ts field is a struct list_head, and is currently initialized in the open() method of the demux device. When for a given demuxer a section filter is set up, the sec field is played with, thus if a PES filter needs to be set up after that the ts field will be corrupted, causing a kernel oops.
This fix moves the list head initialization to dvb_dmxdev_pes_filter_set(), so that the ts field is properly initialized every time a PES filter is set up.

---
 drivers/media/dvb/dvb-core/dmxdev.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/media/dvb/dvb-core/dmxdev.c +1 -1
@@ -761,7 +761,6 @@ static int dvb_demux_open(struct inode *
 	dvb_ringbuffer_init(&dmxdevfilter->buffer, NULL, 8192);
 	dmxdevfilter->type = DMXDEV_TYPE_NONE;
 	dvb_dmxdev_filter_state_set(dmxdevfilter, DMXDEV_STATE_ALLOCATED);
-	INIT_LIST_HEAD(&dmxdevfilter->feed.ts);
 	init_timer(&dmxdevfilter->timer);
 
 	dvbdev->users++;
@@ -887,6 +886,7 @@ static int dvb_dmxdev_pes_filter_set(str
 	dmxdevfilter->type = DMXDEV_TYPE_PES;
 	memcpy(&dmxdevfilter->params, params,
 	       sizeof(struct dmx_pes_filter_params));
+	INIT_LIST_HEAD(&dmxdevfilter->feed.ts);
 
 	dvb_dmxdev_filter_state_set(dmxdevfilter, DMXDEV_STATE_SET);
 
author
#44 lore raw

PATCH 41/93 Export the symbol of getboottime and mmonotonic_to_bootbased

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Jason Wang <[email protected]>

commit c93d89f3dbf0202bf19c07960ca8602b48c2f9a0 upstream.

Export getboottime and monotonic_to_bootbased in order to let them could be used by following patch.

---
 kernel/time/timekeeping.c |    2 ++
 1 file changed, 2 insertions(+)
kernel/time/timekeeping.c +2 -0
@@ -845,6 +845,7 @@ void getboottime(struct timespec *ts)
 
 	set_normalized_timespec(ts, -boottime.tv_sec, -boottime.tv_nsec);
 }
+EXPORT_SYMBOL_GPL(getboottime);
 
 /**
  * monotonic_to_bootbased - Convert the monotonic time to boot based.
@@ -854,6 +855,7 @@ void monotonic_to_bootbased(struct times
 {
 	*ts = timespec_add_safe(*ts, total_sleep_time);
 }
+EXPORT_SYMBOL_GPL(monotonic_to_bootbased);
 
 unsigned long get_seconds(void)
 {
author
#45 lore raw

PATCH 42/93 kvmclock: count total_sleep_time when updating guest clock

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Jason Wang <[email protected]>

commit 923de3cf5bf12049628019010e36623fca5ef6d1 upstream.

Current kvm wallclock does not consider the total_sleep_time which could cause wrong wallclock in guest after host suspend/resume. This patch solve this issue by counting total_sleep_time to get the correct host boot time.

---
 arch/x86/kvm/x86.c |    7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)
arch/x86/kvm/x86.c +3 -4
@@ -583,7 +583,7 @@ static void kvm_write_wall_clock(struct
 {
 	static int version;
 	struct pvclock_wall_clock wc;
-	struct timespec now, sys, boot;
+	struct timespec boot;
 
 	if (!wall_clock)
 		return;
@@ -598,9 +598,7 @@ static void kvm_write_wall_clock(struct
 	 * wall clock specified here.  guest system time equals host
 	 * system time for us, thus we must fill in host boot time here.
 	 */
-	now = current_kernel_time();
-	ktime_get_ts(&sys);
-	boot = ns_to_timespec(timespec_to_ns(&now) - timespec_to_ns(&sys));
+	getboottime(&boot);
 
 	wc.sec = boot.tv_sec;
 	wc.nsec = boot.tv_nsec;
@@ -675,6 +673,7 @@ static void kvm_write_guest_time(struct
 	local_irq_save(flags);
 	kvm_get_msr(v, MSR_IA32_TSC, &vcpu->hv_clock.tsc_timestamp);
 	ktime_get_ts(&ts);
+	monotonic_to_bootbased(&ts);
 	local_irq_restore(flags);
 
 	/* With all the info we got, fill in the values */
author
#46 lore raw

PATCH 43/93 KVM: PIT: control word is write-only

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Marcelo Tosatti <[email protected]>

commit ee73f656a604d5aa9df86a97102e4e462dd79924 upstream.

PIT control word (address 0x43) is write-only, reads are undefined.

---
 arch/x86/kvm/i8254.c |    3 +++
 1 file changed, 3 insertions(+)
arch/x86/kvm/i8254.c +3 -0
@@ -465,6 +465,9 @@ static int pit_ioport_read(struct kvm_io
 		return -EOPNOTSUPP;
 
 	addr &= KVM_PIT_CHANNEL_MASK;
+	if (addr == 3)
+		return 0;
+
 	s = &pit_state->channels[addr];
 
 	mutex_lock(&pit_state->lock);
author
#47 lore raw

PATCH 44/93 tpm_infineon: fix suspend/resume handler for pnp_driver

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Marcel Selhorst <[email protected]>

commit 93716b9470fbfd9efdc7d0f2445cb34635de3f6d upstream.

When suspending, tpm_infineon calls the generic suspend function of the TPM framework. However, the TPM framework does not return and the system hangs upon suspend. When sending the necessary command "TPM_SaveState" directly within the driver, suspending and resuming works fine.

---
 drivers/char/tpm/tpm_infineon.c |   79 ++++++++++++++++++++++++++++------------
 1 file changed, 57 insertions(+), 22 deletions(-)
drivers/char/tpm/tpm_infineon.c +57 -22
@@ -39,12 +39,12 @@
 struct tpm_inf_dev {
 	int iotype;
 
-	void __iomem *mem_base;		/* MMIO ioremap'd addr */
-	unsigned long map_base;		/* phys MMIO base */
-	unsigned long map_size;		/* MMIO region size */
-	unsigned int index_off;		/* index register offset */
+	void __iomem *mem_base;	/* MMIO ioremap'd addr */
+	unsigned long map_base;	/* phys MMIO base */
+	unsigned long map_size;	/* MMIO region size */
+	unsigned int index_off;	/* index register offset */
 
-	unsigned int data_regs;		/* Data registers */
+	unsigned int data_regs;	/* Data registers */
 	unsigned int data_size;
 
 	unsigned int config_port;	/* IO Port config index reg */
@@ -406,14 +406,14 @@ static const struct tpm_vendor_specific
 	.miscdev = {.fops = &inf_ops,},
 };
 
-static const struct pnp_device_id tpm_pnp_tbl[] = {
+static const struct pnp_device_id tpm_inf_pnp_tbl[] = {
 	/* Infineon TPMs */
 	{"IFX0101", 0},
 	{"IFX0102", 0},
 	{"", 0}
 };
 
-MODULE_DEVICE_TABLE(pnp, tpm_pnp_tbl);
+MODULE_DEVICE_TABLE(pnp, tpm_inf_pnp_tbl);
 
 static int __devinit tpm_inf_pnp_probe(struct pnp_dev *dev,
 				       const struct pnp_device_id *dev_id)
@@ -430,7 +430,7 @@ static int __devinit tpm_inf_pnp_probe(s
 	if (pnp_port_valid(dev, 0) && pnp_port_valid(dev, 1) &&
 	    !(pnp_port_flags(dev, 0) & IORESOURCE_DISABLED)) {
 
-	    	tpm_dev.iotype = TPM_INF_IO_PORT;
+		tpm_dev.iotype = TPM_INF_IO_PORT;
 
 		tpm_dev.config_port = pnp_port_start(dev, 0);
 		tpm_dev.config_size = pnp_port_len(dev, 0);
@@ -459,9 +459,9 @@ static int __devinit tpm_inf_pnp_probe(s
 			goto err_last;
 		}
 	} else if (pnp_mem_valid(dev, 0) &&
-	           !(pnp_mem_flags(dev, 0) & IORESOURCE_DISABLED)) {
+		   !(pnp_mem_flags(dev, 0) & IORESOURCE_DISABLED)) {
 
-	    	tpm_dev.iotype = TPM_INF_IO_MEM;
+		tpm_dev.iotype = TPM_INF_IO_MEM;
 
 		tpm_dev.map_base = pnp_mem_start(dev, 0);
 		tpm_dev.map_size = pnp_mem_len(dev, 0);
@@ -563,11 +563,11 @@ static int __devinit tpm_inf_pnp_probe(s
 			 "product id 0x%02x%02x"
 			 "%s\n",
 			 tpm_dev.iotype == TPM_INF_IO_PORT ?
-				tpm_dev.config_port :
-				tpm_dev.map_base + tpm_dev.index_off,
+			 tpm_dev.config_port :
+			 tpm_dev.map_base + tpm_dev.index_off,
 			 tpm_dev.iotype == TPM_INF_IO_PORT ?
-				tpm_dev.data_regs :
-				tpm_dev.map_base + tpm_dev.data_regs,
+			 tpm_dev.data_regs :
+			 tpm_dev.map_base + tpm_dev.data_regs,
 			 version[0], version[1],
 			 vendorid[0], vendorid[1],
 			 productid[0], productid[1], chipname);
@@ -607,20 +607,55 @@ static __devexit void tpm_inf_pnp_remove
 			iounmap(tpm_dev.mem_base);
 			release_mem_region(tpm_dev.map_base, tpm_dev.map_size);
 		}
+		tpm_dev_vendor_release(chip);
 		tpm_remove_hardware(chip->dev);
 	}
 }
 
+static int tpm_inf_pnp_suspend(struct pnp_dev *dev, pm_message_t pm_state)
+{
+	struct tpm_chip *chip = pnp_get_drvdata(dev);
+	int rc;
+	if (chip) {
+		u8 savestate[] = {
+			0, 193,	/* TPM_TAG_RQU_COMMAND */
+			0, 0, 0, 10,	/* blob length (in bytes) */
+			0, 0, 0, 152	/* TPM_ORD_SaveState */
+		};
+		dev_info(&dev->dev, "saving TPM state\n");
+		rc = tpm_inf_send(chip, savestate, sizeof(savestate));
+		if (rc < 0) {
+			dev_err(&dev->dev, "error while saving TPM state\n");
+			return rc;
+		}
+	}
+	return 0;
+}
+
+static int tpm_inf_pnp_resume(struct pnp_dev *dev)
+{
+	/* Re-configure TPM after suspending */
+	tpm_config_out(ENABLE_REGISTER_PAIR, TPM_INF_ADDR);
+	tpm_config_out(IOLIMH, TPM_INF_ADDR);
+	tpm_config_out((tpm_dev.data_regs >> 8) & 0xff, TPM_INF_DATA);
+	tpm_config_out(IOLIML, TPM_INF_ADDR);
+	tpm_config_out((tpm_dev.data_regs & 0xff), TPM_INF_DATA);
+	/* activate register */
+	tpm_config_out(TPM_DAR, TPM_INF_ADDR);
+	tpm_config_out(0x01, TPM_INF_DATA);
+	tpm_config_out(DISABLE_REGISTER_PAIR, TPM_INF_ADDR);
+	/* disable RESET, LP and IRQC */
+	tpm_data_out(RESET_LP_IRQC_DISABLE, CMD);
+	return tpm_pm_resume(&dev->dev);
+}
+
 static struct pnp_driver tpm_inf_pnp_driver = {
 	.name = "tpm_inf_pnp",
-	.driver = {
-		.owner = THIS_MODULE,
-		.suspend = tpm_pm_suspend,
-		.resume = tpm_pm_resume,
-	},
-	.id_table = tpm_pnp_tbl,
+	.id_table = tpm_inf_pnp_tbl,
 	.probe = tpm_inf_pnp_probe,
-	.remove = __devexit_p(tpm_inf_pnp_remove),
+	.suspend = tpm_inf_pnp_suspend,
+	.resume = tpm_inf_pnp_resume,
+	.remove = __devexit_p(tpm_inf_pnp_remove)
 };
 
 static int __init init_inf(void)
@@ -638,5 +673,5 @@ module_exit(cleanup_inf);
 
 MODULE_AUTHOR("Marcel Selhorst <[email protected]>");
 MODULE_DESCRIPTION("Driver for Infineon TPM SLD 9630 TT 1.1 / SLB 9635 TT 1.2");
-MODULE_VERSION("1.9");
+MODULE_VERSION("1.9.2");
 MODULE_LICENSE("GPL");
author
#48 lore raw

PATCH 45/93 amd64_edac: Do not falsely trigger kerneloops

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Borislav Petkov <[email protected]>

commit cab4d27764d5a8654212b3e96eb0ae793aec5b94 upstream.

An unfortunate "WARNING" in the message amd64_edac dumps when the system doesn't support DRAM ECC or ECC checking is not enabled in the BIOS used to trigger kerneloops which qualified the message as an OOPS thus misleading the users. See, e.g.

https://bugs.launchpad.net/ubuntu/+source/linux/+bug/422536
http://bugzilla.kernel.org/show_bug.cgi?id=15238

Downgrade the message level to KERN_NOTICE and fix the formulation.

---
 drivers/edac/amd64_edac.c |   15 ++++++++-------
 1 file changed, 8 insertions(+), 7 deletions(-)
drivers/edac/amd64_edac.c +8 -7
@@ -2801,10 +2801,11 @@ static void amd64_restore_ecc_error_repo
  * the memory system completely. A command line option allows to force-enable
  * hardware ECC later in amd64_enable_ecc_error_reporting().
  */
-static const char *ecc_warning =
-	"WARNING: ECC is disabled by BIOS. Module will NOT be loaded.\n"
-	" Either Enable ECC in the BIOS, or set 'ecc_enable_override'.\n"
-	" Also, use of the override can cause unknown side effects.\n";
+static const char *ecc_msg =
+	"ECC disabled in the BIOS or no ECC capability, module will not load.\n"
+	" Either enable ECC checking or force module loading by setting "
+	"'ecc_enable_override'.\n"
+	" (Note that use of the override may cause unknown side effects.)\n";
 
 static int amd64_check_ecc_enabled(struct amd64_pvt *pvt)
 {
@@ -2819,7 +2820,7 @@ static int amd64_check_ecc_enabled(struc
 
 	ecc_enabled = !!(value & K8_NBCFG_ECC_ENABLE);
 	if (!ecc_enabled)
-		amd64_printk(KERN_WARNING, "This node reports that Memory ECC "
+		amd64_printk(KERN_NOTICE, "This node reports that Memory ECC "
 			     "is currently disabled, set F3x%x[22] (%s).\n",
 			     K8_NBCFG, pci_name(pvt->misc_f3_ctl));
 	else
@@ -2827,13 +2828,13 @@ static int amd64_check_ecc_enabled(struc
 
 	nb_mce_en = amd64_nb_mce_bank_enabled_on_node(pvt->mc_node_id);
 	if (!nb_mce_en)
-		amd64_printk(KERN_WARNING, "NB MCE bank disabled, set MSR "
+		amd64_printk(KERN_NOTICE, "NB MCE bank disabled, set MSR "
 			     "0x%08x[4] on node %d to enable.\n",
 			     MSR_IA32_MCG_CTL, pvt->mc_node_id);
 
 	if (!ecc_enabled || !nb_mce_en) {
 		if (!ecc_enable_override) {
-			amd64_printk(KERN_WARNING, "%s", ecc_warning);
+			amd64_printk(KERN_NOTICE, "%s", ecc_msg);
 			return -ENODEV;
 		}
 		ecc_enable_override = 0;
author
#49 lore raw

PATCH 46/93 netfilter: nf_conntrack: fix memory corruption with multiple namespaces

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Patrick McHardy <[email protected]>

commit 9edd7ca0a3e3999c260642c92fa008892d82ca6e upstream.

As discovered by Jon Masters <[email protected]>, the "untracked" conntrack, which is located in the data section, might be accidentally freed when a new namespace is instantiated while the untracked conntrack is attached to a skb because the reference count it re-initialized.

The best fix would be to use a seperate untracked conntrack per namespace since it includes a namespace pointer. Unfortunately this is not possible without larger changes since the namespace is not easily available everywhere we need it. For now move the untracked conntrack initialization to the init_net setup function to make sure the reference count is not re-initialized and handle cleanup in the init_net cleanup function to make sure namespaces can exit properly while the untracked conntrack is in use in other namespaces.

---
 net/netfilter/nf_conntrack_core.c |   24 ++++++++++++------------
 1 file changed, 12 insertions(+), 12 deletions(-)
net/netfilter/nf_conntrack_core.c +12 -12
@@ -1107,6 +1107,10 @@ static void nf_ct_release_dying_list(str
 
 static void nf_conntrack_cleanup_init_net(void)
 {
+	/* wait until all references to nf_conntrack_untracked are dropped */
+	while (atomic_read(&nf_conntrack_untracked.ct_general.use) > 1)
+		schedule();
+
 	nf_conntrack_helper_fini();
 	nf_conntrack_proto_fini();
 	kmem_cache_destroy(nf_conntrack_cachep);
@@ -1121,9 +1125,6 @@ static void nf_conntrack_cleanup_net(str
 		schedule();
 		goto i_see_dead_people;
 	}
-	/* wait until all references to nf_conntrack_untracked are dropped */
-	while (atomic_read(&nf_conntrack_untracked.ct_general.use) > 1)
-		schedule();
 
 	nf_ct_free_hashtable(net->ct.hash, net->ct.hash_vmalloc,
 			     nf_conntrack_htable_size);
@@ -1282,6 +1283,14 @@ static int nf_conntrack_init_init_net(vo
 	if (ret < 0)
 		goto err_helper;
 
+	/* Set up fake conntrack: to never be deleted, not in any hashes */
+#ifdef CONFIG_NET_NS
+	nf_conntrack_untracked.ct_net = &init_net;
+#endif
+	atomic_set(&nf_conntrack_untracked.ct_general.use, 1);
+	/*  - and look it like as a confirmed connection */
+	set_bit(IPS_CONFIRMED_BIT, &nf_conntrack_untracked.status);
+
 	return 0;
 
 err_helper:
@@ -1327,15 +1336,6 @@ static int nf_conntrack_init_net(struct
 	if (ret < 0)
 		goto err_ecache;
 
-	/* Set up fake conntrack:
-	    - to never be deleted, not in any hashes */
-#ifdef CONFIG_NET_NS
-	nf_conntrack_untracked.ct_net = &init_net;
-#endif
-	atomic_set(&nf_conntrack_untracked.ct_general.use, 1);
-	/*  - and look it like as a confirmed connection */
-	set_bit(IPS_CONFIRMED_BIT, &nf_conntrack_untracked.status);
-
 	return 0;
 
 err_ecache:
author
#50 lore raw

PATCH 47/93 netfilter: nf_conntrack: per netns nf_conntrack_cachep

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Eric Dumazet <[email protected]>

commit 5b3501faa8741d50617ce4191c20061c6ef36cb3 upstream.

nf_conntrack_cachep is currently shared by all netns instances, but because of SLAB_DESTROY_BY_RCU special semantics, this is wrong.

If we use a shared slab cache, one object can instantly flight between one hash table (netns ONE) to another one (netns TWO), and concurrent reader (doing a lookup in netns ONE, 'finding' an object of netns TWO) can be fooled without notice, because no RCU grace period has to be observed between object freeing and its reuse.

We dont have this problem with UDP/TCP slab caches because TCP/UDP hashtables are global to the machine (and each object has a pointer to its netns).

If we use per netns conntrack hash tables, we also *must* use per netns conntrack slab caches, to guarantee an object can not escape from one namespace to another one.

[Patrick: added unique slab name allocation]
---
 include/net/netns/conntrack.h     |    2 +
 net/netfilter/nf_conntrack_core.c |   39 ++++++++++++++++++++++----------------
 2 files changed, 25 insertions(+), 16 deletions(-)
include/net/netns/conntrack.h +2 -0
@@ -11,6 +11,7 @@ struct nf_conntrack_ecache;
 struct netns_ct {
 	atomic_t		count;
 	unsigned int		expect_count;
+	struct kmem_cache	*nf_conntrack_cachep;
 	struct hlist_nulls_head	*hash;
 	struct hlist_head	*expect_hash;
 	struct hlist_nulls_head	unconfirmed;
@@ -28,5 +29,6 @@ struct netns_ct {
 #endif
 	int			hash_vmalloc;
 	int			expect_vmalloc;
+	char			*slabname;
 };
 #endif
net/netfilter/nf_conntrack_core.c +23 -16
@@ -63,8 +63,6 @@ EXPORT_SYMBOL_GPL(nf_conntrack_max);
 struct nf_conn nf_conntrack_untracked __read_mostly;
 EXPORT_SYMBOL_GPL(nf_conntrack_untracked);
 
-static struct kmem_cache *nf_conntrack_cachep __read_mostly;
-
 static int nf_conntrack_hash_rnd_initted;
 static unsigned int nf_conntrack_hash_rnd;
 
@@ -566,7 +564,7 @@ struct nf_conn *nf_conntrack_alloc(struc
 	 * Do not use kmem_cache_zalloc(), as this cache uses
 	 * SLAB_DESTROY_BY_RCU.
 	 */
-	ct = kmem_cache_alloc(nf_conntrack_cachep, gfp);
+	ct = kmem_cache_alloc(net->ct.nf_conntrack_cachep, gfp);
 	if (ct == NULL) {
 		pr_debug("nf_conntrack_alloc: Can't alloc conntrack.\n");
 		atomic_dec(&net->ct.count);
@@ -605,7 +603,7 @@ void nf_conntrack_free(struct nf_conn *c
 	nf_ct_ext_destroy(ct);
 	atomic_dec(&net->ct.count);
 	nf_ct_ext_free(ct);
-	kmem_cache_free(nf_conntrack_cachep, ct);
+	kmem_cache_free(net->ct.nf_conntrack_cachep, ct);
 }
 EXPORT_SYMBOL_GPL(nf_conntrack_free);
 
@@ -1113,7 +1111,6 @@ static void nf_conntrack_cleanup_init_ne
 
 	nf_conntrack_helper_fini();
 	nf_conntrack_proto_fini();
-	kmem_cache_destroy(nf_conntrack_cachep);
 }
 
 static void nf_conntrack_cleanup_net(struct net *net)
@@ -1131,6 +1128,8 @@ static void nf_conntrack_cleanup_net(str
 	nf_conntrack_ecache_fini(net);
 	nf_conntrack_acct_fini(net);
 	nf_conntrack_expect_fini(net);
+	kmem_cache_destroy(net->ct.nf_conntrack_cachep);
+	kfree(net->ct.slabname);
 	free_percpu(net->ct.stat);
 }
 
@@ -1266,15 +1265,6 @@ static int nf_conntrack_init_init_net(vo
 	       NF_CONNTRACK_VERSION, nf_conntrack_htable_size,
 	       nf_conntrack_max);
 
-	nf_conntrack_cachep = kmem_cache_create("nf_conntrack",
-						sizeof(struct nf_conn),
-						0, SLAB_DESTROY_BY_RCU, NULL);
-	if (!nf_conntrack_cachep) {
-		printk(KERN_ERR "Unable to create nf_conn slab cache\n");
-		ret = -ENOMEM;
-		goto err_cache;
-	}
-
 	ret = nf_conntrack_proto_init();
 	if (ret < 0)
 		goto err_proto;
@@ -1296,8 +1286,6 @@ static int nf_conntrack_init_init_net(vo
 err_helper:
 	nf_conntrack_proto_fini();
 err_proto:
-	kmem_cache_destroy(nf_conntrack_cachep);
-err_cache:
 	return ret;
 }
 
@@ -1319,6 +1307,21 @@ static int nf_conntrack_init_net(struct
 		ret = -ENOMEM;
 		goto err_stat;
 	}
+
+	net->ct.slabname = kasprintf(GFP_KERNEL, "nf_conntrack_%p", net);
+	if (!net->ct.slabname) {
+		ret = -ENOMEM;
+		goto err_slabname;
+	}
+
+	net->ct.nf_conntrack_cachep = kmem_cache_create(net->ct.slabname,
+							sizeof(struct nf_conn), 0,
+							SLAB_DESTROY_BY_RCU, NULL);
+	if (!net->ct.nf_conntrack_cachep) {
+		printk(KERN_ERR "Unable to create nf_conn slab cache\n");
+		ret = -ENOMEM;
+		goto err_cache;
+	}
 	net->ct.hash = nf_ct_alloc_hashtable(&nf_conntrack_htable_size,
 					     &net->ct.hash_vmalloc, 1);
 	if (!net->ct.hash) {
@@ -1346,6 +1349,10 @@ err_expect:
 	nf_ct_free_hashtable(net->ct.hash, net->ct.hash_vmalloc,
 			     nf_conntrack_htable_size);
 err_hash:
+	kmem_cache_destroy(net->ct.nf_conntrack_cachep);
+err_cache:
+	kfree(net->ct.slabname);
+err_slabname:
 	free_percpu(net->ct.stat);
 err_stat:
 	return ret;
author
#51 lore raw

PATCH 48/93 netfilter: nf_conntrack: restrict runtime expect hashsize modifications

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Alexey Dobriyan <[email protected]>

commit 13ccdfc2af03e09e60791f7d4bc4ccf53398af7c upstream.

Expectation hashtable size was simply glued to a variable with no code to rehash expectations, so it was a bug to allow writing to it. Make "expect_hashsize" readonly.

---
 net/netfilter/nf_conntrack_expect.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
net/netfilter/nf_conntrack_expect.c +1 -1
@@ -569,7 +569,7 @@ static void exp_proc_remove(struct net *
 #endif /* CONFIG_PROC_FS */
 }
 
-module_param_named(expect_hashsize, nf_ct_expect_hsize, uint, 0600);
+module_param_named(expect_hashsize, nf_ct_expect_hsize, uint, 0400);
 
 int nf_conntrack_expect_init(struct net *net)
 {
author
#52 lore raw

PATCH 49/93 netfilter: xtables: compat out of scope fix

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Alexey Dobriyan <[email protected]>

commit 14c7dbe043d01a83a30633ab6b109ba2ac61d9f7 upstream.

As per C99 6.2.4(2) when temporary table data goes out of scope, the behaviour is undefined:

	if (compat) {
		struct foo tmp;
		...
		private = &tmp;
	}
	[dereference private]
---
 net/ipv4/netfilter/arp_tables.c |    4 ++--
 net/ipv4/netfilter/ip_tables.c  |    4 ++--
 net/ipv6/netfilter/ip6_tables.c |    4 ++--
 3 files changed, 6 insertions(+), 6 deletions(-)
net/ipv4/netfilter/arp_tables.c +2 -2
@@ -925,10 +925,10 @@ static int get_info(struct net *net, voi
 	if (t && !IS_ERR(t)) {
 		struct arpt_getinfo info;
 		const struct xt_table_info *private = t->private;
-
 #ifdef CONFIG_COMPAT
+		struct xt_table_info tmp;
+
 		if (compat) {
-			struct xt_table_info tmp;
 			ret = compat_table_info(private, &tmp);
 			xt_compat_flush_offsets(NFPROTO_ARP);
 			private = &tmp;
net/ipv4/netfilter/ip_tables.c +2 -2
@@ -1132,10 +1132,10 @@ static int get_info(struct net *net, voi
 	if (t && !IS_ERR(t)) {
 		struct ipt_getinfo info;
 		const struct xt_table_info *private = t->private;
-
 #ifdef CONFIG_COMPAT
+		struct xt_table_info tmp;
+
 		if (compat) {
-			struct xt_table_info tmp;
 			ret = compat_table_info(private, &tmp);
 			xt_compat_flush_offsets(AF_INET);
 			private = &tmp;
net/ipv6/netfilter/ip6_tables.c +2 -2
@@ -1164,10 +1164,10 @@ static int get_info(struct net *net, voi
 	if (t && !IS_ERR(t)) {
 		struct ip6t_getinfo info;
 		const struct xt_table_info *private = t->private;
-
 #ifdef CONFIG_COMPAT
+		struct xt_table_info tmp;
+
 		if (compat) {
-			struct xt_table_info tmp;
 			ret = compat_table_info(private, &tmp);
 			xt_compat_flush_offsets(AF_INET6);
 			private = &tmp;
author
#53 lore raw

PATCH 50/93 netfilter: nf_conntrack: fix hash resizing with namespaces

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Patrick McHardy <[email protected]>

commit d696c7bdaa55e2208e56c6f98e6bc1599f34286d upstream.

As noticed by Jon Masters <[email protected]>, the conntrack hash size is global and not per namespace, but modifiable at runtime through /sys/module/nf_conntrack/hashsize. Changing the hash size will only resize the hash in the current namespace however, so other namespaces will use an invalid hash size. This can cause crashes when enlarging the hashsize, or false negative lookups when shrinking it.

Move the hash size into the per-namespace data and only use the global hash size to initialize the per-namespace value when instanciating a new namespace. Additionally restrict hash resizing to init_net for now as other namespaces are not handled currently.

---
 include/net/netns/conntrack.h                         |    1 
 include/net/netns/ipv4.h                              |    1 
 net/ipv4/netfilter/nf_conntrack_l3proto_ipv4.c        |    2 
 net/ipv4/netfilter/nf_conntrack_l3proto_ipv4_compat.c |    4 -
 net/ipv4/netfilter/nf_nat_core.c                      |   22 +++----
 net/netfilter/nf_conntrack_core.c                     |   54 +++++++++---------
 net/netfilter/nf_conntrack_expect.c                   |    2 
 net/netfilter/nf_conntrack_helper.c                   |    2 
 net/netfilter/nf_conntrack_netlink.c                  |    2 
 net/netfilter/nf_conntrack_standalone.c               |    7 +-
 10 files changed, 50 insertions(+), 47 deletions(-)
include/net/netns/conntrack.h +1 -0
@@ -11,6 +11,7 @@ struct nf_conntrack_ecache;
 struct netns_ct {
 	atomic_t		count;
 	unsigned int		expect_count;
+	unsigned int		htable_size;
 	struct kmem_cache	*nf_conntrack_cachep;
 	struct hlist_nulls_head	*hash;
 	struct hlist_head	*expect_hash;
include/net/netns/ipv4.h +1 -0
@@ -40,6 +40,7 @@ struct netns_ipv4 {
 	struct xt_table		*iptable_security;
 	struct xt_table		*nat_table;
 	struct hlist_head	*nat_bysource;
+	unsigned int		nat_htable_size;
 	int			nat_vmalloced;
 #endif
 
net/ipv4/netfilter/nf_conntrack_l3proto_ipv4.c +1 -1
@@ -213,7 +213,7 @@ static ctl_table ip_ct_sysctl_table[] =
 	{
 		.ctl_name	= NET_IPV4_NF_CONNTRACK_BUCKETS,
 		.procname	= "ip_conntrack_buckets",
-		.data		= &nf_conntrack_htable_size,
+		.data		= &init_net.ct.htable_size,
 		.maxlen		= sizeof(unsigned int),
 		.mode		= 0444,
 		.proc_handler	= proc_dointvec,
net/ipv4/netfilter/nf_conntrack_l3proto_ipv4_compat.c +2 -2
@@ -32,7 +32,7 @@ static struct hlist_nulls_node *ct_get_f
 	struct hlist_nulls_node *n;
 
 	for (st->bucket = 0;
-	     st->bucket < nf_conntrack_htable_size;
+	     st->bucket < net->ct.htable_size;
 	     st->bucket++) {
 		n = rcu_dereference(net->ct.hash[st->bucket].first);
 		if (!is_a_nulls(n))
@@ -50,7 +50,7 @@ static struct hlist_nulls_node *ct_get_n
 	head = rcu_dereference(head->next);
 	while (is_a_nulls(head)) {
 		if (likely(get_nulls_value(head) == st->bucket)) {
-			if (++st->bucket >= nf_conntrack_htable_size)
+			if (++st->bucket >= net->ct.htable_size)
 				return NULL;
 		}
 		head = rcu_dereference(net->ct.hash[st->bucket].first);
net/ipv4/netfilter/nf_nat_core.c +9 -13
@@ -35,9 +35,6 @@ static DEFINE_SPINLOCK(nf_nat_lock);
 
 static struct nf_conntrack_l3proto *l3proto __read_mostly;
 
-/* Calculated at init based on memory size */
-static unsigned int nf_nat_htable_size __read_mostly;
-
 #define MAX_IP_NAT_PROTO 256
 static const struct nf_nat_protocol *nf_nat_protos[MAX_IP_NAT_PROTO]
 						__read_mostly;
@@ -72,7 +69,7 @@ EXPORT_SYMBOL_GPL(nf_nat_proto_put);
 
 /* We keep an extra hash for each conntrack, for fast searching. */
 static inline unsigned int
-hash_by_src(const struct nf_conntrack_tuple *tuple)
+hash_by_src(const struct net *net, const struct nf_conntrack_tuple *tuple)
 {
 	unsigned int hash;
 
@@ -80,7 +77,7 @@ hash_by_src(const struct nf_conntrack_tu
 	hash = jhash_3words((__force u32)tuple->src.u3.ip,
 			    (__force u32)tuple->src.u.all,
 			    tuple->dst.protonum, 0);
-	return ((u64)hash * nf_nat_htable_size) >> 32;
+	return ((u64)hash * net->ipv4.nat_htable_size) >> 32;
 }
 
 /* Is this tuple already taken? (not by us) */
@@ -147,7 +144,7 @@ find_appropriate_src(struct net *net,
 		     struct nf_conntrack_tuple *result,
 		     const struct nf_nat_range *range)
 {
-	unsigned int h = hash_by_src(tuple);
+	unsigned int h = hash_by_src(net, tuple);
 	const struct nf_conn_nat *nat;
 	const struct nf_conn *ct;
 	const struct hlist_node *n;
@@ -330,7 +327,7 @@ nf_nat_setup_info(struct nf_conn *ct,
 	if (have_to_hash) {
 		unsigned int srchash;
 
-		srchash = hash_by_src(&ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple);
+		srchash = hash_by_src(net, &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple);
 		spin_lock_bh(&nf_nat_lock);
 		/* nf_conntrack_alter_reply might re-allocate exntension aera */
 		nat = nfct_nat(ct);
@@ -679,8 +676,10 @@ nfnetlink_parse_nat_setup(struct nf_conn
 
 static int __net_init nf_nat_net_init(struct net *net)
 {
-	net->ipv4.nat_bysource = nf_ct_alloc_hashtable(&nf_nat_htable_size,
-						      &net->ipv4.nat_vmalloced, 0);
+	/* Leave them the same for the moment. */
+	net->ipv4.nat_htable_size = net->ct.htable_size;
+	net->ipv4.nat_bysource = nf_ct_alloc_hashtable(&net->ipv4.nat_htable_size,
+						       &net->ipv4.nat_vmalloced, 0);
 	if (!net->ipv4.nat_bysource)
 		return -ENOMEM;
 	return 0;
@@ -703,7 +702,7 @@ static void __net_exit nf_nat_net_exit(s
 	nf_ct_iterate_cleanup(net, &clean_nat, NULL);
 	synchronize_rcu();
 	nf_ct_free_hashtable(net->ipv4.nat_bysource, net->ipv4.nat_vmalloced,
-			     nf_nat_htable_size);
+			     net->ipv4.nat_htable_size);
 }
 
 static struct pernet_operations nf_nat_net_ops = {
@@ -724,9 +723,6 @@ static int __init nf_nat_init(void)
 		return ret;
 	}
 
-	/* Leave them the same for the moment. */
-	nf_nat_htable_size = nf_conntrack_htable_size;
-
 	ret = register_pernet_subsys(&nf_nat_net_ops);
 	if (ret < 0)
 		goto cleanup_extend;
net/netfilter/nf_conntrack_core.c +29 -25

168 lines not on this page · raw message

net/netfilter/nf_conntrack_expect.c +1 -1
@@ -577,7 +577,7 @@ int nf_conntrack_expect_init(struct net
 
 	if (net_eq(net, &init_net)) {
 		if (!nf_ct_expect_hsize) {
-			nf_ct_expect_hsize = nf_conntrack_htable_size / 256;
+			nf_ct_expect_hsize = net->ct.htable_size / 256;
 			if (!nf_ct_expect_hsize)
 				nf_ct_expect_hsize = 1;
 		}
net/netfilter/nf_conntrack_helper.c +1 -1
@@ -192,7 +192,7 @@ static void __nf_conntrack_helper_unregi
 	/* Get rid of expecteds, set helpers to NULL. */
 	hlist_nulls_for_each_entry(h, nn, &net->ct.unconfirmed, hnnode)
 		unhelp(h, me);
-	for (i = 0; i < nf_conntrack_htable_size; i++) {
+	for (i = 0; i < net->ct.htable_size; i++) {
 		hlist_nulls_for_each_entry(h, nn, &net->ct.hash[i], hnnode)
 			unhelp(h, me);
 	}
net/netfilter/nf_conntrack_netlink.c +1 -1
@@ -594,7 +594,7 @@ ctnetlink_dump_table(struct sk_buff *skb
 
 	rcu_read_lock();
 	last = (struct nf_conn *)cb->args[1];
-	for (; cb->args[0] < nf_conntrack_htable_size; cb->args[0]++) {
+	for (; cb->args[0] < init_net.ct.htable_size; cb->args[0]++) {
 restart:
 		hlist_nulls_for_each_entry_rcu(h, n, &init_net.ct.hash[cb->args[0]],
 					 hnnode) {
net/netfilter/nf_conntrack_standalone.c +4 -3
@@ -51,7 +51,7 @@ static struct hlist_nulls_node *ct_get_f
 	struct hlist_nulls_node *n;
 
 	for (st->bucket = 0;
-	     st->bucket < nf_conntrack_htable_size;
+	     st->bucket < net->ct.htable_size;
 	     st->bucket++) {
 		n = rcu_dereference(net->ct.hash[st->bucket].first);
 		if (!is_a_nulls(n))
@@ -69,7 +69,7 @@ static struct hlist_nulls_node *ct_get_n
 	head = rcu_dereference(head->next);
 	while (is_a_nulls(head)) {
 		if (likely(get_nulls_value(head) == st->bucket)) {
-			if (++st->bucket >= nf_conntrack_htable_size)
+			if (++st->bucket >= net->ct.htable_size)
 				return NULL;
 		}
 		head = rcu_dereference(net->ct.hash[st->bucket].first);
@@ -358,7 +358,7 @@ static ctl_table nf_ct_sysctl_table[] =
 	{
 		.ctl_name       = NET_NF_CONNTRACK_BUCKETS,
 		.procname       = "nf_conntrack_buckets",
-		.data           = &nf_conntrack_htable_size,
+		.data           = &init_net.ct.htable_size,
 		.maxlen         = sizeof(unsigned int),
 		.mode           = 0444,
 		.proc_handler   = proc_dointvec,
@@ -429,6 +429,7 @@ static int nf_conntrack_standalone_init_
 		goto out_kmemdup;
 
 	table[1].data = &net->ct.count;
+	table[2].data = &net->ct.htable_size;
 	table[3].data = &net->ct.sysctl_checksum;
 	table[4].data = &net->ct.sysctl_log_invalid;
 
author
#54 lore raw

PATCH 51/93 drm/i915: remove full registers dump debug

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Zhenyu Wang <[email protected]>

commit 823f68fd646da6a39a9c0d3eb4c60d69dab5aa13 upstream.

This one reverts 9e3a6d155ed0a7636b926a798dd7221ea107b274. As reported by http://bugzilla.kernel.org/show_bug.cgi?id=14485,
this dump will cause hang problem on some machine. If something really needs this kind of full registers dump, that could be done within intel-gpu-tools.

---
 drivers/gpu/drm/i915/i915_debugfs.c |   30 ------------------------------
 1 file changed, 30 deletions(-)
drivers/gpu/drm/i915/i915_debugfs.c +0 -30
@@ -384,37 +384,7 @@ out:
 	return 0;
 }
 
-static int i915_registers_info(struct seq_file *m, void *data) {
-	struct drm_info_node *node = (struct drm_info_node *) m->private;
-	struct drm_device *dev = node->minor->dev;
-	drm_i915_private_t *dev_priv = dev->dev_private;
-	uint32_t reg;
-
-#define DUMP_RANGE(start, end) \
-	for (reg=start; reg < end; reg += 4) \
-	seq_printf(m, "%08x\t%08x\n", reg, I915_READ(reg));
-
-	DUMP_RANGE(0x00000, 0x00fff);   /* VGA registers */
-	DUMP_RANGE(0x02000, 0x02fff);   /* instruction, memory, interrupt control registers */
-	DUMP_RANGE(0x03000, 0x031ff);   /* FENCE and PPGTT control registers */
-	DUMP_RANGE(0x03200, 0x03fff);   /* frame buffer compression registers */
-	DUMP_RANGE(0x05000, 0x05fff);   /* I/O control registers */
-	DUMP_RANGE(0x06000, 0x06fff);   /* clock control registers */
-	DUMP_RANGE(0x07000, 0x07fff);   /* 3D internal debug registers */
-	DUMP_RANGE(0x07400, 0x088ff);   /* GPE debug registers */
-	DUMP_RANGE(0x0a000, 0x0afff);   /* display palette registers */
-	DUMP_RANGE(0x10000, 0x13fff);   /* MMIO MCHBAR */
-	DUMP_RANGE(0x30000, 0x3ffff);   /* overlay registers */
-	DUMP_RANGE(0x60000, 0x6ffff);   /* display engine pipeline registers */
-	DUMP_RANGE(0x70000, 0x72fff);   /* display and cursor registers */
-	DUMP_RANGE(0x73000, 0x73fff);   /* performance counters */
-
-	return 0;
-}
-
-
 static struct drm_info_list i915_debugfs_list[] = {
-	{"i915_regs", i915_registers_info, 0},
 	{"i915_gem_active", i915_gem_object_list_info, 0, (void *) ACTIVE_LIST},
 	{"i915_gem_flushing", i915_gem_object_list_info, 0, (void *) FLUSHING_LIST},
 	{"i915_gem_inactive", i915_gem_object_list_info, 0, (void *) INACTIVE_LIST},
author
#55 lore raw

PATCH 52/93 drm/i915: add i915_lp_ring_sync helper

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Daniel Vetter <[email protected]>

commit 48764bf43f746113fc77877d7e80f2df23ca4cbb upstream.

This just waits until the hw passed the current ring position with cmd execution. This slightly changes the existing i915_wait_request function to make uninterruptible waiting possible - no point in returning to userspace while mucking around with the overlay, that piece of hw is just too fragile.

Also replace a magic 0 with the symbolic constant (and kill the then superflous comment) while I was looking at the code.

---
 drivers/gpu/drm/i915/i915_drv.h |    1 
 drivers/gpu/drm/i915/i915_gem.c |   49 +++++++++++++++++++++++++++++++---------
 include/drm/drm_os_linux.h      |    2 -
 3 files changed, 41 insertions(+), 11 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +1 -0
@@ -825,6 +825,7 @@ void i915_gem_cleanup_ringbuffer(struct
 int i915_gem_do_init(struct drm_device *dev, unsigned long start,
 		     unsigned long end);
 int i915_gem_idle(struct drm_device *dev);
+int i915_lp_ring_sync(struct drm_device *dev);
 int i915_gem_fault(struct vm_area_struct *vma, struct vm_fault *vmf);
 int i915_gem_object_set_to_gtt_domain(struct drm_gem_object *obj,
 				      int write);
drivers/gpu/drm/i915/i915_gem.c +39 -10

74 lines not on this page · raw message

include/drm/drm_os_linux.h +1 -1
@@ -123,5 +123,5 @@ do {								\
 	remove_wait_queue(&(queue), &entry);			\
 } while (0)
 
-#define DRM_WAKEUP( queue ) wake_up_interruptible( queue )
+#define DRM_WAKEUP( queue ) wake_up( queue )
 #define DRM_INIT_WAITQUEUE( queue ) init_waitqueue_head( queue )
author
#56 lore raw

PATCH 53/93 drm/i915: Dont wait interruptible for possible plane buffer flush

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Zhenyu Wang <[email protected]>

commit b9241ea31fae4887104e5d1b3b18f4009c25a0c4 upstream.

When we setup buffer for display plane, we'll check any pending required GPU flush and possible make interruptible wait for flush complete. But that wait would be most possibly to fail in case of signals received for X process, which will then fail modeset process and put display engine in unconsistent state. The result could be blank screen or CPU hang, and DDX driver would always turn on outputs DPMS after whatever modeset fails or not.

So this one creates new helper for setup display plane buffer, and when needing flush using uninterruptible wait for that.

This one should fix bug like https://bugs.freedesktop.org/show_bug.cgi?id=24009.
Also fixing mode switch stress test on Ironlake.

---
 drivers/gpu/drm/i915/i915_drv.h      |    1 
 drivers/gpu/drm/i915/i915_gem.c      |   51 +++++++++++++++++++++++++++++++++++
 drivers/gpu/drm/i915/intel_display.c |    2 -
 3 files changed, 53 insertions(+), 1 deletion(-)
drivers/gpu/drm/i915/i915_drv.h +1 -0
@@ -829,6 +829,7 @@ int i915_lp_ring_sync(struct drm_device
 int i915_gem_fault(struct vm_area_struct *vma, struct vm_fault *vmf);
 int i915_gem_object_set_to_gtt_domain(struct drm_gem_object *obj,
 				      int write);
+int i915_gem_object_set_to_display_plane(struct drm_gem_object *obj);
 int i915_gem_attach_phys_object(struct drm_device *dev,
 				struct drm_gem_object *obj, int id);
 void i915_gem_detach_phys_object(struct drm_device *dev,
drivers/gpu/drm/i915/i915_gem.c +51 -0

57 lines not on this page · raw message

drivers/gpu/drm/i915/intel_display.c +1 -1
@@ -1253,7 +1253,7 @@ intel_pipe_set_base(struct drm_crtc *crt
 		return ret;
 	}
 
-	ret = i915_gem_object_set_to_gtt_domain(obj, 1);
+	ret = i915_gem_object_set_to_display_plane(obj);
 	if (ret != 0) {
 		i915_gem_object_unpin(obj);
 		mutex_unlock(&dev->struct_mutex);
author
#57 lore raw

PATCH 54/93 dasd: remove strings from s390dbf

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Stefan Haberland <[email protected]>

commit b8ed5dd54895647c2690575aad6f07748c2c618a upstream.

Remove strings from s390 debugfeature entries that could lead to a crash when the data is read from dbf because the strings do not exist any more.

---
 drivers/s390/block/dasd.c      |   22 ++++++++++----------
 drivers/s390/block/dasd_eckd.c |   44 +++++++++++++++++------------------------
 drivers/s390/block/dasd_fba.c  |   10 +++------
 drivers/s390/block/dasd_int.h  |   10 +++++++++
 4 files changed, 44 insertions(+), 42 deletions(-)
drivers/s390/block/dasd.c +11 -11
@@ -994,10 +994,9 @@ static void dasd_handle_killed_request(s
 		return;
 	cqr = (struct dasd_ccw_req *) intparm;
 	if (cqr->status != DASD_CQR_IN_IO) {
-		DBF_EVENT(DBF_DEBUG,
-			"invalid status in handle_killed_request: "
-			"bus_id %s, status %02x",
-			dev_name(&cdev->dev), cqr->status);
+		DBF_EVENT_DEVID(DBF_DEBUG, cdev,
+				"invalid status in handle_killed_request: "
+				"%02x", cqr->status);
 		return;
 	}
 
@@ -1045,12 +1044,13 @@ void dasd_int_handler(struct ccw_device
 		case -EIO:
 			break;
 		case -ETIMEDOUT:
-			DBF_EVENT(DBF_WARNING, "%s(%s): request timed out\n",
-			       __func__, dev_name(&cdev->dev));
+			DBF_EVENT_DEVID(DBF_WARNING, cdev, "%s: "
+					"request timed out\n", __func__);
 			break;
 		default:
-			DBF_EVENT(DBF_WARNING, "%s(%s): unknown error %ld\n",
-			       __func__, dev_name(&cdev->dev), PTR_ERR(irb));
+			DBF_EVENT_DEVID(DBF_WARNING, cdev, "%s: "
+					"unknown error %ld\n", __func__,
+					PTR_ERR(irb));
 		}
 		dasd_handle_killed_request(cdev, intparm);
 		return;
@@ -2217,9 +2217,9 @@ int dasd_generic_probe(struct ccw_device
 	}
 	ret = dasd_add_sysfs_files(cdev);
 	if (ret) {
-		DBF_EVENT(DBF_WARNING,
-		       "dasd_generic_probe: could not add sysfs entries "
-		       "for %s\n", dev_name(&cdev->dev));
+		DBF_EVENT_DEVID(DBF_WARNING, cdev, "%s",
+				"dasd_generic_probe: could not add "
+				"sysfs entries");
 		return ret;
 	}
 	cdev->handler = &dasd_int_handler;
drivers/s390/block/dasd_eckd.c +19 -25

90 lines not on this page · raw message

drivers/s390/block/dasd_fba.c +4 -6
@@ -141,9 +141,8 @@ dasd_fba_check_characteristics(struct da
 	}
 	block = dasd_alloc_block();
 	if (IS_ERR(block)) {
-		DBF_EVENT(DBF_WARNING, "could not allocate dasd block "
-			  "structure for device: %s",
-			  dev_name(&device->cdev->dev));
+		DBF_EVENT_DEVID(DBF_WARNING, cdev, "%s", "could not allocate "
+				"dasd block structure");
 		device->private = NULL;
 		kfree(private);
 		return PTR_ERR(block);
@@ -155,9 +154,8 @@ dasd_fba_check_characteristics(struct da
 	rc = dasd_generic_read_dev_chars(device, DASD_FBA_MAGIC,
 					 &private->rdc_data, 32);
 	if (rc) {
-		DBF_EVENT(DBF_WARNING, "Read device characteristics returned "
-			  "error %d for device: %s",
-			  rc, dev_name(&device->cdev->dev));
+		DBF_EVENT_DEVID(DBF_WARNING, cdev, "Read device "
+				"characteristics returned error %d", rc);
 		device->block = NULL;
 		dasd_free_block(block);
 		device->private = NULL;
drivers/s390/block/dasd_int.h +10 -0
@@ -108,6 +108,16 @@ do { \
 			    d_data); \
 } while(0)
 
+#define DBF_EVENT_DEVID(d_level, d_cdev, d_str, d_data...)	\
+do { \
+	struct ccw_dev_id __dev_id;			\
+	ccw_device_get_id(d_cdev, &__dev_id);		\
+	debug_sprintf_event(dasd_debug_area,		\
+			    d_level,					\
+			    "0.%x.%04x " d_str "\n",			\
+			    __dev_id.ssid, __dev_id.devno, d_data);	\
+} while (0)
+
 #define DBF_EXC(d_level, d_str, d_data...)\
 do { \
 	debug_sprintf_exception(dasd_debug_area, \
author
#58 lore raw

PATCH 55/93 crypto: padlock-sha - Add import/export support

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Herbert Xu <[email protected]>

commit a8d7ac279743077965afeca0c9ed748507b68e89 upstream.

As the padlock driver for SHA uses a software fallback to perform partial hashing, it must implement custom import/export functions. Otherwise hmac which depends on import/export for prehashing will not work with padlock-sha.

---
 drivers/crypto/padlock-sha.c |   23 +++++++++++++++++++++++
 1 file changed, 23 insertions(+)
drivers/crypto/padlock-sha.c +23 -0
@@ -57,6 +57,23 @@ static int padlock_sha_update(struct sha
 	return crypto_shash_update(&dctx->fallback, data, length);
 }
 
+static int padlock_sha_export(struct shash_desc *desc, void *out)
+{
+	struct padlock_sha_desc *dctx = shash_desc_ctx(desc);
+
+	return crypto_shash_export(&dctx->fallback, out);
+}
+
+static int padlock_sha_import(struct shash_desc *desc, const void *in)
+{
+	struct padlock_sha_desc *dctx = shash_desc_ctx(desc);
+	struct padlock_sha_ctx *ctx = crypto_shash_ctx(desc->tfm);
+
+	dctx->fallback.tfm = ctx->fallback;
+	dctx->fallback.flags = desc->flags & CRYPTO_TFM_REQ_MAY_SLEEP;
+	return crypto_shash_import(&dctx->fallback, in);
+}
+
 static inline void padlock_output_block(uint32_t *src,
 		 	uint32_t *dst, size_t count)
 {
@@ -235,7 +252,10 @@ static struct shash_alg sha1_alg = {
 	.update 	=	padlock_sha_update,
 	.finup  	=	padlock_sha1_finup,
 	.final  	=	padlock_sha1_final,
+	.export		=	padlock_sha_export,
+	.import		=	padlock_sha_import,
 	.descsize	=	sizeof(struct padlock_sha_desc),
+	.statesize	=	sizeof(struct sha1_state),
 	.base		=	{
 		.cra_name		=	"sha1",
 		.cra_driver_name	=	"sha1-padlock",
@@ -256,7 +276,10 @@ static struct shash_alg sha256_alg = {
 	.update 	=	padlock_sha_update,
 	.finup  	=	padlock_sha256_finup,
 	.final  	=	padlock_sha256_final,
+	.export		=	padlock_sha_export,
+	.import		=	padlock_sha_import,
 	.descsize	=	sizeof(struct padlock_sha_desc),
+	.statesize	=	sizeof(struct sha256_state),
 	.base		=	{
 		.cra_name		=	"sha256",
 		.cra_driver_name	=	"sha256-padlock",
author
#59 lore raw

PATCH 56/93 wmi: Free the allocated acpi objects through wmi_get_event_data

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Anisse Astier <[email protected]>

commit 3e9b988e4edf065d39c1343937f717319b1c1065 upstream

[ backported to 2.6.32 ]

These function allocate an acpi object by calling wmi_get_event_data, which then calls acpi_evaluate_object, and it is not freed afterwards.

And kernel doc is fixed for parameters of wmi_get_event_data.

---
 drivers/platform/x86/dell-wmi.c |    1 +
 drivers/platform/x86/hp-wmi.c   |    2 ++
 drivers/platform/x86/wmi.c      |    4 ++--
 3 files changed, 5 insertions(+), 2 deletions(-)
drivers/platform/x86/dell-wmi.c +1 -0
@@ -180,6 +180,7 @@ static void dell_wmi_notify(u32 value, v
 			printk(KERN_INFO "dell-wmi: Unknown key %x pressed\n",
 			       buffer[1] & 0xFFFF);
 	}
+	kfree(obj);
 }
 
 static int __init dell_wmi_input_setup(void)
drivers/platform/x86/hp-wmi.c +2 -0
@@ -377,6 +377,8 @@ static void hp_wmi_notify(u32 value, voi
 			       eventcode);
 	} else
 		printk(KERN_INFO "HP WMI: Unknown response received\n");
+
+	kfree(obj);
 }
 
 static int __init hp_wmi_input_setup(void)
drivers/platform/x86/wmi.c +2 -2
@@ -510,8 +510,8 @@ EXPORT_SYMBOL_GPL(wmi_remove_notify_hand
 /**
  * wmi_get_event_data - Get WMI data associated with an event
  *
- * @event - Event to find
- * &out - Buffer to hold event data
+ * @event: Event to find
+ * @out: Buffer to hold event data. out->pointer should be freed with kfree()
  *
  * Returns extra data associated with an event in WMI.
  */
author
#60 lore raw

PATCH 57/93 dell-wmi, hp-wmi, msi-wmi: check wmi_get_event_data() return value

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Len Brown <[email protected]>

commit fda11e61ff8a4e3a8ebbd434e46560b67cc0ca9d upstream

[ backport to 2.6.32 ]

When acpi_evaluate_object() is passed ACPI_ALLOCATE_BUFFER, the caller must kfree the returned buffer if AE_OK is returned.

The callers of wmi_get_event_data() pass ACPI_ALLOCATE_BUFFER, and thus must check its return value before accessing or kfree() on the buffer.

---
 drivers/platform/x86/dell-wmi.c |    7 ++++++-
 drivers/platform/x86/hp-wmi.c   |    7 ++++++-
 2 files changed, 12 insertions(+), 2 deletions(-)
drivers/platform/x86/dell-wmi.c +6 -1
@@ -158,8 +158,13 @@ static void dell_wmi_notify(u32 value, v
 	struct acpi_buffer response = { ACPI_ALLOCATE_BUFFER, NULL };
 	static struct key_entry *key;
 	union acpi_object *obj;
+	acpi_status status;
 
-	wmi_get_event_data(value, &response);
+	status = wmi_get_event_data(value, &response);
+	if (status != AE_OK) {
+		printk(KERN_INFO "dell-wmi: bad event status 0x%x\n", status);
+		return;
+	}
 
 	obj = (union acpi_object *)response.pointer;
 
drivers/platform/x86/hp-wmi.c +6 -1
@@ -334,8 +334,13 @@ static void hp_wmi_notify(u32 value, voi
 	struct acpi_buffer response = { ACPI_ALLOCATE_BUFFER, NULL };
 	static struct key_entry *key;
 	union acpi_object *obj;
+	acpi_status status;
 
-	wmi_get_event_data(value, &response);
+	status = wmi_get_event_data(value, &response);
+	if (status != AE_OK) {
+		printk(KERN_INFO "hp-wmi: bad event status 0x%x\n", status);
+		return;
+	}
 
 	obj = (union acpi_object *)response.pointer;
 
author
#61 lore raw

PATCH 58/93 /dev/mem: introduce size_inside_page()

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Wu Fengguang <[email protected]>

commit f222318e9c3a315723e3524fb9d6566b2430db44 upstream

/dev/mem: introduce size_inside_page()

[ [email protected] : backport to 2.6.32 ]
[ subset of original patch, for just /dev/kmem ]

Introduce size_inside_page() to replace duplicate /dev/mem code.

Also apply it to /dev/kmem, whose alignment logic was buggy.

---
 drivers/char/mem.c |   40 +++++++++++++++++-----------------------
 1 file changed, 17 insertions(+), 23 deletions(-)
drivers/char/mem.c +17 -23
@@ -35,6 +35,19 @@
 # include <linux/efi.h>
 #endif
 
+static inline unsigned long size_inside_page(unsigned long start,
+					     unsigned long size)
+{
+	unsigned long sz;
+
+	if (-start & (PAGE_SIZE - 1))
+		sz = -start & (PAGE_SIZE - 1);
+	else
+		sz = PAGE_SIZE;
+
+	return min_t(unsigned long, sz, size);
+}
+
 /*
  * Architectures vary in how they handle caching for addresses
  * outside of main memory.
@@ -430,15 +443,7 @@ static ssize_t read_kmem(struct file *fi
 		}
 #endif
 		while (low_count > 0) {
-			/*
-			 * Handle first page in case it's not aligned
-			 */
-			if (-p & (PAGE_SIZE - 1))
-				sz = -p & (PAGE_SIZE - 1);
-			else
-				sz = PAGE_SIZE;
-
-			sz = min_t(unsigned long, sz, low_count);
+			sz = size_inside_page(p, low_count);
 
 			/*
 			 * On ia64 if a page has been mapped somewhere as
@@ -462,10 +467,8 @@ static ssize_t read_kmem(struct file *fi
 		if (!kbuf)
 			return -ENOMEM;
 		while (count > 0) {
-			int len = count;
+			int len = size_inside_page(p, count);
 
-			if (len > PAGE_SIZE)
-				len = PAGE_SIZE;
 			len = vread(kbuf, (char *)p, len);
 			if (!len)
 				break;
@@ -510,15 +513,8 @@ do_write_kmem(void *p, unsigned long rea
 
 	while (count > 0) {
 		char *ptr;
-		/*
-		 * Handle first page in case it's not aligned
-		 */
-		if (-realp & (PAGE_SIZE - 1))
-			sz = -realp & (PAGE_SIZE - 1);
-		else
-			sz = PAGE_SIZE;
 
-		sz = min_t(unsigned long, sz, count);
+		sz = size_inside_page(realp, count);
 
 		/*
 		 * On ia64 if a page has been mapped somewhere as
@@ -578,10 +574,8 @@ static ssize_t write_kmem(struct file *
 		if (!kbuf)
 			return wrote ? wrote : -ENOMEM;
 		while (count > 0) {
-			int len = count;
+			int len = size_inside_page(p, count);
 
-			if (len > PAGE_SIZE)
-				len = PAGE_SIZE;
 			if (len) {
 				written = copy_from_user(kbuf, buf, len);
 				if (written) {
author
#62 lore raw

PATCH 59/93 devmem: check vmalloc address on kmem read/write

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: KAMEZAWA Hiroyuki <[email protected]>

commit 325fda71d0badc1073dc59f12a948f24ff05796a

[ [email protected] : backport to 2.6.32 ]

devmem: check vmalloc address on kmem read/write

Otherwise vmalloc_to_page() will BUG().

This also makes the kmem read/write implementation aligned with mem(4):
"References to nonexistent locations cause errors to be returned." Here we return -ENXIO (inspired by Hugh) if no bytes have been transfered to/from user space, otherwise return partial read/write results.

---
 drivers/char/mem.c |   28 ++++++++++++++++++----------
 1 file changed, 18 insertions(+), 10 deletions(-)
drivers/char/mem.c +18 -10
@@ -421,6 +421,7 @@ static ssize_t read_kmem(struct file *fi
 	unsigned long p = *ppos;
 	ssize_t low_count, read, sz;
 	char * kbuf; /* k-addr because vread() takes vmlist_lock rwlock */
+	int err = 0;
 
 	read = 0;
 	if (p < (unsigned long) high_memory) {
@@ -469,12 +470,16 @@ static ssize_t read_kmem(struct file *fi
 		while (count > 0) {
 			int len = size_inside_page(p, count);
 
+			if (!is_vmalloc_or_module_addr((void *)p)) {
+				err = -ENXIO;
+				break;
+			}
 			len = vread(kbuf, (char *)p, len);
 			if (!len)
 				break;
 			if (copy_to_user(buf, kbuf, len)) {
-				free_page((unsigned long)kbuf);
-				return -EFAULT;
+				err = -EFAULT;
+				break;
 			}
 			count -= len;
 			buf += len;
@@ -483,8 +488,8 @@ static ssize_t read_kmem(struct file *fi
 		}
 		free_page((unsigned long)kbuf);
 	}
- 	*ppos = p;
- 	return read;
+	*ppos = p;
+	return read ? read : err;
 }
 
 
@@ -553,6 +558,7 @@ static ssize_t write_kmem(struct file *
 	ssize_t virtr = 0;
 	ssize_t written;
 	char * kbuf; /* k-addr because vwrite() takes vmlist_lock rwlock */
+	int err = 0;
 
 	if (p < (unsigned long) high_memory) {
 
@@ -576,13 +582,15 @@ static ssize_t write_kmem(struct file *
 		while (count > 0) {
 			int len = size_inside_page(p, count);
 
+			if (!is_vmalloc_or_module_addr((void *)p)) {
+				err = -ENXIO;
+				break;
+			}
 			if (len) {
 				written = copy_from_user(kbuf, buf, len);
 				if (written) {
-					if (wrote + virtr)
-						break;
-					free_page((unsigned long)kbuf);
-					return -EFAULT;
+					err = -EFAULT;
+					break;
 				}
 			}
 			len = vwrite(kbuf, (char *)p, len);
@@ -594,8 +602,8 @@ static ssize_t write_kmem(struct file *
 		free_page((unsigned long)kbuf);
 	}
 
- 	*ppos = p;
- 	return virtr + wrote;
+	*ppos = p;
+	return virtr + wrote ? : err;
 }
 #endif
 
author
#63 lore raw

PATCH 60/93 devmem: fix kmem write bug on memory holes

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Wu Fengguang <[email protected]>

commit c85e9a97c4102ce2e83112da850d838cfab5ab13 upstream

devmem: fix kmem write bug on memory holes

[ [email protected] : backport to 2.6.32 ]

write_kmem() used to assume vwrite() always return the full buffer length. However now vwrite() could return 0 to indicate memory hole. This creates a bug that "buf" is not advanced accordingly.

Fix it to simply ignore the return value, hence the memory hole.

Cc Andi Kleen <[email protected]>
Cc Benjamin Herrenschmidt <[email protected]>
Cc Christoph Lameter <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Tejun Heo <[email protected]>
Cc Nick Piggin <[email protected]>
Cc KAMEZAWA Hiroyuki <[email protected]>
Cc Chuck Ebbert <[email protected]>
---
 drivers/char/mem.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/char/mem.c +1 -1
@@ -593,7 +593,7 @@ static ssize_t write_kmem(struct file *
 					break;
 				}
 			}
-			len = vwrite(kbuf, (char *)p, len);
+			vwrite(kbuf, (char *)p, len);
 			count -= len;
 			buf += len;
 			virtr += len;
author
#64 lore raw

PATCH 61/93 SCSI: mptfusion : mptscsih_abort return value should be SUCCESS instead of value 0.

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Kashyap, Desai <[email protected]>

commit 9858ae38011d699d4c2fa7f3493a47accf43a0f5 upstream.

retval should be SUCCESS/FAILED which is defined at scsi.h retval = 0 is directing wrong return value. It must be retval = SUCCESS.

---
 drivers/message/fusion/mptscsih.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/message/fusion/mptscsih.c +1 -1
@@ -1720,7 +1720,7 @@ mptscsih_abort(struct scsi_cmnd * SCpnt)
 		dtmprintk(ioc, printk(MYIOC_s_DEBUG_FMT "task abort: "
 		   "Command not in the active list! (sc=%p)\n", ioc->name,
 		   SCpnt));
-		retval = 0;
+		retval = SUCCESS;
 		goto out;
 	}
 
author
#65 lore raw

PATCH 62/93 sh: Couple kernel and user write page perm bits for CONFIG_X2TLB

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Matt Fleming <[email protected]>

commit fcb4ebd678858850e8b029909064175cb627868d upstream.

pte_write() should check whether the permissions include either the user or kernel write permission bits. Likewise, pte_wrprotect() needs to remove both the kernel and user write bits.

Without this patch handle_tlbmiss() doesn't handle faulting in pages from the P3 area (our vmalloc space) because of a write. Mappings of the P3 space have the _PAGE_EXT_KERN_WRITE bit but not _PAGE_EXT_USER_WRITE.

---
 arch/sh/include/asm/pgtable_32.h |    5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
arch/sh/include/asm/pgtable_32.h +3 -2
@@ -344,7 +344,8 @@ static inline void set_pte(pte_t *ptep,
 #define pte_special(pte)	((pte).pte_low & _PAGE_SPECIAL)
 
 #ifdef CONFIG_X2TLB
-#define pte_write(pte)		((pte).pte_high & _PAGE_EXT_USER_WRITE)
+#define pte_write(pte) \
+	((pte).pte_high & (_PAGE_EXT_USER_WRITE | _PAGE_EXT_KERN_WRITE))
 #else
 #define pte_write(pte)		((pte).pte_low & _PAGE_RW)
 #endif
@@ -358,7 +359,7 @@ static inline pte_t pte_##fn(pte_t pte)
  * individually toggled (and user permissions are entirely decoupled from
  * kernel permissions), we attempt to couple them a bit more sanely here.
  */
-PTE_BIT_FUNC(high, wrprotect, &= ~_PAGE_EXT_USER_WRITE);
+PTE_BIT_FUNC(high, wrprotect, &= ~(_PAGE_EXT_USER_WRITE | _PAGE_EXT_KERN_WRITE));
 PTE_BIT_FUNC(high, mkwrite, |= _PAGE_EXT_USER_WRITE | _PAGE_EXT_KERN_WRITE);
 PTE_BIT_FUNC(high, mkhuge, |= _PAGE_SZHUGE);
 #else
author
#66 lore raw

PATCH 63/93 ALSA: hda - use WARN_ON_ONCE() for zero-division detection

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Takashi Iwai <[email protected]>

commit d6d8bf549393484e906913f02fa3c9518a2819b6 upstream.

Replace the zero-division warning message with WARN_ON_ONCE() per the advice by Linus. This shouldn't happen, but if it happens, it's possible that the bug happens often due to buggy IRQs.

---
 sound/pci/hda/hda_intel.c |    9 +++------
 1 file changed, 3 insertions(+), 6 deletions(-)
sound/pci/hda/hda_intel.c +3 -6
@@ -1858,12 +1858,9 @@ static int azx_position_ok(struct azx *c
 
 	if (!bdl_pos_adj[chip->dev_index])
 		return 1; /* no delayed ack */
-	if (azx_dev->period_bytes == 0) {
-		printk(KERN_WARNING
-		       "hda-intel: Divide by zero was avoided "
-		       "in azx_dev->period_bytes.\n");
-		return 0;
-	}
+	if (WARN_ONCE(!azx_dev->period_bytes,
+		      "hda-intel: zero azx_dev->period_bytes"))
+		return 0; /* this shouldn't happen! */
 	if (pos % azx_dev->period_bytes > azx_dev->period_bytes / 2)
 		return 0; /* NG - it's below the period boundary */
 	return 1; /* OK, it's fine */
author
#67 lore raw

PATCH 64/93 dst: call cond_resched() in dst_gc_task()

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Eric Dumazet <[email protected]>

commit 2fc1b5dd99f66d93ffc23fd8df82d384c1a354c8 upstream.

Kernel bugzilla #15239

On some workloads, it is quite possible to get a huge dst list to process in dst_gc_task(), and trigger soft lockup detection.

Fix is to call cond_resched(), as we run in process context.

---
 net/core/dst.c |    2 ++
 1 file changed, 2 insertions(+)
net/core/dst.c +2 -0
@@ -17,6 +17,7 @@
 #include <linux/string.h>
 #include <linux/types.h>
 #include <net/net_namespace.h>
+#include <linux/sched.h>
 
 #include <net/dst.h>
 
@@ -79,6 +80,7 @@ loop:
 	while ((dst = next) != NULL) {
 		next = dst->next;
 		prefetch(&next->next);
+		cond_resched();
 		if (likely(atomic_read(&dst->__refcnt))) {
 			last->next = dst;
 			last = dst;
author
#68 lore raw

PATCH 65/93 ALSA: hda - Improved MacBook (Pro) 5,1 / 5,2 support

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Alex Murray <[email protected]>

commit a76221d47ef2b73ff16c0fef00a784026308ea02 upstream.

This patch adds support for automatically muting the speakers when headphones are inserted, as well as relabelling the headphone widgets from the non-standard "HP" to the standard "Headphone" for the mb5 model.

---
 sound/pci/hda/patch_realtek.c |   28 ++++++++++++++++++++++++++--
 1 file changed, 26 insertions(+), 2 deletions(-)
sound/pci/hda/patch_realtek.c +26 -2
@@ -7042,8 +7042,8 @@ static struct snd_kcontrol_new alc885_mb
 	HDA_BIND_MUTE   ("Surround Playback Switch", 0x0d, 0x02, HDA_INPUT),
 	HDA_CODEC_VOLUME("LFE Playback Volume", 0x0e, 0x00, HDA_OUTPUT),
 	HDA_BIND_MUTE   ("LFE Playback Switch", 0x0e, 0x02, HDA_INPUT),
-	HDA_CODEC_VOLUME("HP Playback Volume", 0x0f, 0x00, HDA_OUTPUT),
-	HDA_BIND_MUTE   ("HP Playback Switch", 0x0f, 0x02, HDA_INPUT),
+	HDA_CODEC_VOLUME("Headphone Playback Volume", 0x0f, 0x00, HDA_OUTPUT),
+	HDA_BIND_MUTE   ("Headphone Playback Switch", 0x0f, 0x02, HDA_INPUT),
 	HDA_CODEC_VOLUME("Line Playback Volume", 0x0b, 0x02, HDA_INPUT),
 	HDA_CODEC_MUTE  ("Line Playback Switch", 0x0b, 0x02, HDA_INPUT),
 	HDA_CODEC_VOLUME("Mic Playback Volume", 0x0b, 0x01, HDA_INPUT),
@@ -7430,6 +7430,7 @@ static struct hda_verb alc885_mb5_init_v
 	{0x14, AC_VERB_SET_PIN_WIDGET_CONTROL, PIN_OUT},
 	{0x14, AC_VERB_SET_AMP_GAIN_MUTE, AMP_OUT_UNMUTE},
 	{0x14, AC_VERB_SET_CONNECT_SEL, 0x03},
+	{0x14, AC_VERB_SET_UNSOLICITED_ENABLE, ALC880_HP_EVENT | AC_USRSP_EN},
 	/* Front Mic pin: input vref at 80% */
 	{0x19, AC_VERB_SET_PIN_WIDGET_CONTROL, PIN_VREF80},
 	{0x19, AC_VERB_SET_AMP_GAIN_MUTE, AMP_OUT_MUTE},
@@ -7554,6 +7555,27 @@ static void alc885_mbp3_setup(struct hda
 	spec->autocfg.speaker_pins[0] = 0x14;
 }
 
+static void alc885_mb5_automute(struct hda_codec *codec)
+{
+	unsigned int present;
+
+	present = snd_hda_codec_read(codec, 0x14, 0,
+				     AC_VERB_GET_PIN_SENSE, 0) & 0x80000000;
+	snd_hda_codec_amp_stereo(codec, 0x18, HDA_OUTPUT, 0,
+				 HDA_AMP_MUTE, present ? HDA_AMP_MUTE : 0);
+	snd_hda_codec_amp_stereo(codec, 0x1a, HDA_OUTPUT, 0,
+				 HDA_AMP_MUTE, present ? HDA_AMP_MUTE : 0);
+
+}
+
+static void alc885_mb5_unsol_event(struct hda_codec *codec,
+				    unsigned int res)
+{
+	/* Headphone insertion or removal. */
+	if ((res >> 26) == ALC880_HP_EVENT)
+		alc885_mb5_automute(codec);
+}
+
 
 static struct hda_verb alc882_targa_verbs[] = {
 	{0x0c, AC_VERB_SET_AMP_GAIN_MUTE, AMP_IN_UNMUTE(0)},
@@ -8996,6 +9018,8 @@ static struct alc_config_preset alc882_p
 		.input_mux = &mb5_capture_source,
 		.dig_out_nid = ALC882_DIGOUT_NID,
 		.dig_in_nid = ALC882_DIGIN_NID,
+		.unsol_event = alc885_mb5_unsol_event,
+		.init_hook = alc885_mb5_automute,
 	},
 	[ALC885_MACPRO] = {
 		.mixers = { alc882_macpro_mixer },
author
#69 lore raw

PATCH 66/93 befs: fix leak

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Al Viro <[email protected]>

commit 8dd5ca532c2d2c2b85f16bc038ebfff05b8853e1 upstream.

---
 fs/befs/linuxvfs.c |    1 +
 1 file changed, 1 insertion(+)
fs/befs/linuxvfs.c +1 -0
@@ -873,6 +873,7 @@ befs_fill_super(struct super_block *sb,
 	brelse(bh);
 
       unacquire_priv_sbp:
+	kfree(befs_sb->mount_opts.iocharset);
 	kfree(sb->s_fs_info);
 
       unacquire_none:
author
#70 lore raw

PATCH 67/93 rtc-fm3130: add missing braces

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Sergey Matyukevich <[email protected]>

commit f4b5162820de60204afa5c8639335f4931b7fb0c upstream.

Add missing braces for multiline 'if' statements in fm3130_probe.

drivers/rtc/rtc-fm3130.c +4 -2
@@ -376,20 +376,22 @@ static int __devinit fm3130_probe(struct i2c_client *client,
 	}
 
 	/* Disabling calibration mode */
-	if (fm3130->regs[FM3130_RTC_CONTROL] & FM3130_RTC_CONTROL_BIT_CAL)
+	if (fm3130->regs[FM3130_RTC_CONTROL] & FM3130_RTC_CONTROL_BIT_CAL) {
 		i2c_smbus_write_byte_data(client, FM3130_RTC_CONTROL,
 			fm3130->regs[FM3130_RTC_CONTROL] &
 				~(FM3130_RTC_CONTROL_BIT_CAL));
 		dev_warn(&client->dev, "Disabling calibration mode!\n");
+	}
 
 	/* Disabling read and write modes */
 	if (fm3130->regs[FM3130_RTC_CONTROL] & FM3130_RTC_CONTROL_BIT_WRITE ||
-	    fm3130->regs[FM3130_RTC_CONTROL] & FM3130_RTC_CONTROL_BIT_READ)
+	    fm3130->regs[FM3130_RTC_CONTROL] & FM3130_RTC_CONTROL_BIT_READ) {
 		i2c_smbus_write_byte_data(client, FM3130_RTC_CONTROL,
 			fm3130->regs[FM3130_RTC_CONTROL] &
 				~(FM3130_RTC_CONTROL_BIT_READ |
 					FM3130_RTC_CONTROL_BIT_WRITE));
 		dev_warn(&client->dev, "Disabling READ or WRITE mode!\n");
+	}
 
 	/* oscillator off?  turn it on, so clock can tick. */
 	if (fm3130->regs[FM3130_CAL_CONTROL] & FM3130_CAL_CONTROL_BIT_nOSCEN)
author
#71 lore raw

PATCH 68/93 Call flush_dcache_page after PIO data transfers in libata-sff.c

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Catalin Marinas <[email protected]>

commit 2d68b7fe55d9e19a8a868224ed0dfd6526568521 upstream.

flush_dcache_page() must be called after (!ATA_TFLAG_WRITE) the data copying to avoid D-cache aliasing with user space or I-D cache coherency issues (when reading data from an ATA device using PIO, the kernel dirties the D-cache but there is no flush_dcache_page() required on Harvard architectures).

---
 drivers/ata/libata-sff.c |    3 +++
 1 file changed, 3 insertions(+)
drivers/ata/libata-sff.c +3 -0
@@ -893,6 +893,9 @@ static void ata_pio_sector(struct ata_qu
 				       do_write);
 	}
 
+	if (!do_write)
+		flush_dcache_page(page);
+
 	qc->curbytes += qc->sect_size;
 	qc->cursg_ofs += qc->sect_size;
 
author
#72 lore raw

PATCH 69/93 ahci: add Acer G725 to broken suspend list

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Tejun Heo <[email protected]>

commit cedc9bf906dae044443d403371c887affdb44168 upstream.

Acer G725 shares the same suspend problem with the HP laptops which lose ATA devices on resume. New firmware which fixes the problem is already available. Add G725 with old firmwares to the broken suspend list.

This problem has been reported in bko#15104.

http://bugzilla.kernel.org/show_bug.cgi?id=15104

---
 drivers/ata/ahci.c |   15 +++++++++++++++
 1 file changed, 15 insertions(+)
drivers/ata/ahci.c +15 -0
@@ -2868,6 +2868,21 @@ static bool ahci_broken_suspend(struct p
 			},
 			.driver_data = "F.23",	/* cutoff BIOS version */
 		},
+		/*
+		 * Acer eMachines G725 has the same problem.  BIOS
+		 * V1.03 is known to be broken.  V3.04 is known to
+		 * work.  Inbetween, there are V1.06, V2.06 and V3.03
+		 * that we don't have much idea about.  For now,
+		 * blacklist anything older than V3.04.
+		 */
+		{
+			.ident = "G725",
+			.matches = {
+				DMI_MATCH(DMI_SYS_VENDOR, "eMachines"),
+				DMI_MATCH(DMI_PRODUCT_NAME, "eMachines G725"),
+			},
+			.driver_data = "V3.04",	/* cutoff BIOS version */
+		},
 		{ }	/* terminate list */
 	};
 	const struct dmi_system_id *dmi = dmi_first_match(sysids);
author
#73 lore raw

PATCH 70/93 pktgen: Fix freezing problem

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Rafael J. Wysocki <[email protected]>

commit 1b3f720bf033fde1fbb6231f9b156b918c5f68d8 upstream.

Add missing try_to_freeze() to one of the pktgen_thread_worker() code paths so that it doesn't block suspend/hibernation.

Fixes http://bugzilla.kernel.org/show_bug.cgi?id=15006

---
 net/core/pktgen.c |    1 +
 1 file changed, 1 insertion(+)
net/core/pktgen.c +1 -0
@@ -3516,6 +3516,7 @@ static int pktgen_thread_worker(void *ar
 			wait_event_interruptible_timeout(t->queue,
 							 t->control != 0,
 							 HZ/10);
+			try_to_freeze();
 			continue;
 		}
 
author
#74 lore raw

PATCH 71/93 x86/amd-iommu: Fix IOMMU-API initialization for iommu=pt

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Joerg Roedel <[email protected]>

commit f5325094379158e6b876ea0010c807bf7890ec8f upstream

This patch moves the initialization of the iommu-api out of the dma-ops initialization code. This ensures that the iommu-api is initialized even with iommu=pt.

---
 arch/x86/include/asm/amd_iommu.h |    1 +
 arch/x86/kernel/amd_iommu.c      |    7 +++++--
 arch/x86/kernel/amd_iommu_init.c |    3 +++
 3 files changed, 9 insertions(+), 2 deletions(-)
arch/x86/include/asm/amd_iommu.h +1 -0
@@ -32,6 +32,7 @@ extern void amd_iommu_flush_all_domains(
 extern void amd_iommu_flush_all_devices(void);
 extern void amd_iommu_shutdown(void);
 extern void amd_iommu_apply_erratum_63(u16 devid);
+extern void amd_iommu_init_api(void);
 #else
 static inline int amd_iommu_init(void) { return -ENODEV; }
 static inline void amd_iommu_detect(void) { }
arch/x86/kernel/amd_iommu.c +5 -2
@@ -2083,6 +2083,11 @@ static struct dma_map_ops amd_iommu_dma_
 	.dma_supported = amd_iommu_dma_supported,
 };
 
+void __init amd_iommu_init_api(void)
+{
+	register_iommu(&amd_iommu_ops);
+}
+
 /*
  * The function which clues the AMD IOMMU driver into dma_ops.
  */
@@ -2124,8 +2129,6 @@ int __init amd_iommu_init_dma_ops(void)
 	/* Make the driver finally visible to the drivers */
 	dma_ops = &amd_iommu_dma_ops;
 
-	register_iommu(&amd_iommu_ops);
-
 	bus_register_notifier(&pci_bus_type, &device_nb);
 
 	amd_iommu_stats_init();
arch/x86/kernel/amd_iommu_init.c +3 -0
@@ -1288,9 +1288,12 @@ int __init amd_iommu_init(void)
 		ret = amd_iommu_init_passthrough();
 	else
 		ret = amd_iommu_init_dma_ops();
+
 	if (ret)
 		goto free;
 
+	amd_iommu_init_api();
+
 	enable_iommus();
 
 	if (iommu_pass_through)
author
#75 lore raw

PATCH 72/93 x86/amd-iommu: Fix deassignment of a device from the pt_domain

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Joerg Roedel <[email protected]>

commit d3ad9373b7c29b63d5e8460a69453718d200cc3b upstream.

Deassigning a device from the passthrough domain does not work and breaks device assignment to kvm guests. This patch fixes the issue.

---
 arch/x86/kernel/amd_iommu.c |    5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
arch/x86/kernel/amd_iommu.c +3 -2
@@ -1230,9 +1230,10 @@ static void __detach_device(struct prote
 
 	/*
 	 * If we run in passthrough mode the device must be assigned to the
-	 * passthrough domain if it is detached from any other domain
+	 * passthrough domain if it is detached from any other domain.
+	 * Make sure we can deassign from the pt_domain itself.
 	 */
-	if (iommu_pass_through) {
+	if (iommu_pass_through && domain != pt_domain) {
 		struct amd_iommu *iommu = amd_iommu_rlookup_table[devid];
 		__attach_device(iommu, pt_domain, devid);
 	}
author
#76 lore raw

PATCH 73/93 x86: Re-get cfg_new in case reuse/move irq_desc

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Yinghai Lu <[email protected]>

commit 37ef2a3029fde884808ff1b369677abc7dd9a79a upstream.

When irq_desc is moved, we need to make sure to use the right cfg_new.

LKML-Reference: <[email protected]>

---
 arch/x86/kernel/apic/io_apic.c |    1 +
 1 file changed, 1 insertion(+)
arch/x86/kernel/apic/io_apic.c +1 -0
@@ -3157,6 +3157,7 @@ unsigned int create_irq_nr(unsigned int
 			continue;
 
 		desc_new = move_irq_desc(desc_new, node);
+		cfg_new = desc_new->chip_data;
 
 		if (__assign_irq_vector(new, cfg_new, apic->target_cpus()) == 0)
 			irq = new;
author
#77 lore raw

PATCH 74/93 Staging: fix rtl8187se compilation errors with mac80211

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: George Kadianakis <[email protected]>

commit df574b8ecfb3a84af96229f336a6be88ca4a7055 upstream.

This patch fixes compilation problems that were caused by function naming conflicts between the rtl8187se driver and the mac80211 stack.

---
 drivers/staging/rtl8187se/ieee80211/ieee80211.h         |   10 +++++-----
 drivers/staging/rtl8187se/ieee80211/ieee80211_rx.c      |    2 +-
 drivers/staging/rtl8187se/ieee80211/ieee80211_softmac.c |   14 +++++++-------
 drivers/staging/rtl8187se/ieee80211/ieee80211_tx.c      |    2 +-
 drivers/staging/rtl8187se/r8180_core.c                  |   10 +++++-----
 drivers/staging/rtl8187se/r8180_wx.c                    |    2 +-
 6 files changed, 20 insertions(+), 20 deletions(-)
drivers/staging/rtl8187se/ieee80211/ieee80211.h +5 -5
@@ -1318,13 +1318,13 @@ extern int ieee80211_encrypt_fragment(
 	struct sk_buff *frag,
 	int hdr_len);
 
-extern int ieee80211_xmit(struct sk_buff *skb,
+extern int ieee80211_rtl_xmit(struct sk_buff *skb,
 			  struct net_device *dev);
 extern void ieee80211_txb_free(struct ieee80211_txb *);
 
 
 /* ieee80211_rx.c */
-extern int ieee80211_rx(struct ieee80211_device *ieee, struct sk_buff *skb,
+extern int ieee80211_rtl_rx(struct ieee80211_device *ieee, struct sk_buff *skb,
 			struct ieee80211_rx_stats *rx_stats);
 extern void ieee80211_rx_mgt(struct ieee80211_device *ieee,
 			     struct ieee80211_hdr_4addr *header,
@@ -1376,8 +1376,8 @@ extern void ieee80211_stop_protocol(stru
 extern void ieee80211_softmac_start_protocol(struct ieee80211_device *ieee);
 extern void ieee80211_softmac_stop_protocol(struct ieee80211_device *ieee);
 extern void ieee80211_reset_queue(struct ieee80211_device *ieee);
-extern void ieee80211_wake_queue(struct ieee80211_device *ieee);
-extern void ieee80211_stop_queue(struct ieee80211_device *ieee);
+extern void ieee80211_rtl_wake_queue(struct ieee80211_device *ieee);
+extern void ieee80211_rtl_stop_queue(struct ieee80211_device *ieee);
 extern struct sk_buff *ieee80211_get_beacon(struct ieee80211_device *ieee);
 extern void ieee80211_start_send_beacons(struct ieee80211_device *ieee);
 extern void ieee80211_stop_send_beacons(struct ieee80211_device *ieee);
@@ -1385,7 +1385,7 @@ extern int ieee80211_wpa_supplicant_ioct
 extern void notify_wx_assoc_event(struct ieee80211_device *ieee);
 extern void ieee80211_ps_tx_ack(struct ieee80211_device *ieee, short success);
 extern void SendDisassociation(struct ieee80211_device *ieee,u8* asSta,u8 asRsn);
-extern void ieee80211_start_scan(struct ieee80211_device *ieee);
+extern void ieee80211_rtl_start_scan(struct ieee80211_device *ieee);
 
 //Add for RF power on power off by lizhaoming 080512
 extern void SendDisassociation(struct ieee80211_device *ieee,
drivers/staging/rtl8187se/ieee80211/ieee80211_rx.c +1 -1
@@ -470,7 +470,7 @@ drop:
 /* All received frames are sent to this function. @skb contains the frame in
  * IEEE 802.11 format, i.e., in the format it was sent over air.
  * This function is called only as a tasklet (software IRQ). */
-int ieee80211_rx(struct ieee80211_device *ieee, struct sk_buff *skb,
+int ieee80211_rtl_rx(struct ieee80211_device *ieee, struct sk_buff *skb,
 		 struct ieee80211_rx_stats *rx_stats)
 {
 	struct net_device *dev = ieee->dev;
drivers/staging/rtl8187se/ieee80211/ieee80211_softmac.c +7 -7
@@ -689,7 +689,7 @@ void ieee80211_stop_scan(struct ieee8021
 }
 
 /* called with ieee->lock held */
-void ieee80211_start_scan(struct ieee80211_device *ieee)
+void ieee80211_rtl_start_scan(struct ieee80211_device *ieee)
 {
 	if(IS_DOT11D_ENABLE(ieee) )
 	{
@@ -1196,7 +1196,7 @@ void ieee80211_associate_step1(struct ie
 	}
 }
 
-void ieee80211_auth_challenge(struct ieee80211_device *ieee, u8 *challenge, int chlen)
+void ieee80211_rtl_auth_challenge(struct ieee80211_device *ieee, u8 *challenge, int chlen)
 {
 	u8 *c;
 	struct sk_buff *skb;
@@ -1898,7 +1898,7 @@ associate_complete:
 
 								ieee80211_associate_step2(ieee);
 							}else{
-								ieee80211_auth_challenge(ieee, challenge, chlen);
+								ieee80211_rtl_auth_challenge(ieee, challenge, chlen);
 							}
 						}else{
 							ieee->softmac_stats.rx_auth_rs_err++;
@@ -2047,7 +2047,7 @@ void ieee80211_reset_queue(struct ieee80
 
 }
 
-void ieee80211_wake_queue(struct ieee80211_device *ieee)
+void ieee80211_rtl_wake_queue(struct ieee80211_device *ieee)
 {
 
 	unsigned long flags;
@@ -2089,7 +2089,7 @@ exit :
 }
 
 
-void ieee80211_stop_queue(struct ieee80211_device *ieee)
+void ieee80211_rtl_stop_queue(struct ieee80211_device *ieee)
 {
 	//unsigned long flags;
 	//spin_lock_irqsave(&ieee->lock,flags);
@@ -2301,7 +2301,7 @@ void ieee80211_start_bss(struct ieee8021
 //#else
 	if (ieee->state == IEEE80211_NOLINK){
 		ieee->actscanning = true;
-		ieee80211_start_scan(ieee);
+		ieee80211_rtl_start_scan(ieee);
 	}
 //#endif
 	spin_unlock_irqrestore(&ieee->lock, flags);
@@ -2357,7 +2357,7 @@ void ieee80211_associate_retry_wq(struct
 	if(ieee->state == IEEE80211_NOLINK){
 		ieee->beinretry = false;
 		ieee->actscanning = true;
-		ieee80211_start_scan(ieee);
+		ieee80211_rtl_start_scan(ieee);
 	}
 	//YJ,add,080828, notify os here
 	if(ieee->state == IEEE80211_NOLINK)
drivers/staging/rtl8187se/ieee80211/ieee80211_tx.c +1 -1
@@ -305,7 +305,7 @@ ieee80211_classify(struct sk_buff *skb,
 }
 
 /* SKBs are added to the ieee->tx_queue. */
-int ieee80211_xmit(struct sk_buff *skb,
+int ieee80211_rtl_xmit(struct sk_buff *skb,
 		   struct net_device *dev)
 {
 	struct ieee80211_device *ieee = netdev_priv(dev);
drivers/staging/rtl8187se/r8180_core.c +5 -5
@@ -1830,7 +1830,7 @@ void rtl8180_rx(struct net_device *dev)
 			if(priv->rx_skb->len > 4)
 				skb_trim(priv->rx_skb,priv->rx_skb->len-4);
 #ifndef RX_DONT_PASS_UL
-			if(!ieee80211_rx(priv->ieee80211,
+			if(!ieee80211_rtl_rx(priv->ieee80211,
 					 priv->rx_skb, &stats)){
 #endif // RX_DONT_PASS_UL
 
@@ -1936,11 +1936,11 @@ rate)
 	if (!check_nic_enought_desc(dev, priority)){
 		DMESGW("Error: no descriptor left by previous TX (avail %d) ",
 			get_curr_tx_free_desc(dev, priority));
-		ieee80211_stop_queue(priv->ieee80211);
+		ieee80211_rtl_stop_queue(priv->ieee80211);
 	}
 	rtl8180_tx(dev, skb->data, skb->len, priority, morefrag,0,rate);
 	if (!check_nic_enought_desc(dev, priority))
-		ieee80211_stop_queue(priv->ieee80211);
+		ieee80211_rtl_stop_queue(priv->ieee80211);
 
 	spin_unlock_irqrestore(&priv->tx_lock,flags);
 }
@@ -3846,7 +3846,7 @@ static const struct net_device_ops rtl81
 	.ndo_set_mac_address	= r8180_set_mac_adr,
 	.ndo_validate_addr	= eth_validate_addr,
 	.ndo_change_mtu		= eth_change_mtu,
-	.ndo_start_xmit		= ieee80211_xmit,
+	.ndo_start_xmit		= ieee80211_rtl_xmit,
 };
 
 static int __devinit rtl8180_pci_probe(struct pci_dev *pdev,
@@ -4066,7 +4066,7 @@ void rtl8180_try_wake_queue(struct net_d
 	spin_unlock_irqrestore(&priv->tx_lock,flags);
 
 	if(enough_desc)
-		ieee80211_wake_queue(priv->ieee80211);
+		ieee80211_rtl_wake_queue(priv->ieee80211);
 }
 
 void rtl8180_tx_isr(struct net_device *dev, int pri,short error)
drivers/staging/rtl8187se/r8180_wx.c +1 -1
@@ -377,7 +377,7 @@ static int r8180_wx_set_scan(struct net_
 	//	queue_work(priv->ieee80211->wq, &priv->ieee80211->wx_sync_scan_wq);
 		//printk("start scan============================>\n");
 		ieee80211_softmac_ips_scan_syncro(priv->ieee80211);
-//ieee80211_start_scan(priv->ieee80211);
+//ieee80211_rtl_start_scan(priv->ieee80211);
 		/* intentionally forget to up sem */
 //			up(&priv->ieee80211->wx_sem);
 			ret = 0;
author
#78 lore raw

PATCH 75/93 ALSA: usb-audio - Avoid Oops after disconnect

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Takashi Iwai <[email protected]>

commit 78b8d5d2ee280c463908fd75f3bdf246bcb6ac8d upstream.

As the release of substreams may be done asynchronously from the disconnection, close callback needs to check the shutdown flag before actually accessing the usb interface.

Reference: Novell bnc#505027

	http://bugzilla.novell.com/show_bug.cgi?id=565027
---
 sound/usb/usbaudio.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
sound/usb/usbaudio.c +1 -1
@@ -1936,7 +1936,7 @@ static int snd_usb_pcm_close(struct snd_
 	struct snd_usb_stream *as = snd_pcm_substream_chip(substream);
 	struct snd_usb_substream *subs = &as->substream[direction];
 
-	if (subs->interface >= 0) {
+	if (!as->chip->shutdown && subs->interface >= 0) {
 		usb_set_interface(subs->dev, subs->interface, 0);
 		subs->interface = -1;
 	}
author
#79 lore raw

PATCH 76/93 serial: 8250: add serial transmitter fully empty test

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Dick Hollenbeck <[email protected]>

commit bca476139d2ded86be146dae09b06e22548b67f3 upstream.

When controlling an industrial radio modem it can be necessary to manipulate the handshake lines in order to control the radio modem's transmitter, from userspace.

The transmitter should not be turned off before all characters have been transmitted. serial8250_tx_empty() was reporting that all characters were transmitted before they actually were.

===

Discovered in parallel with more testing and analysis by Kees Schoenmakers as follows:

I ran into an NetMos 9835 serial pci board which behaves a little different than the standard. This type of expansion board is very common.

"Standard" 8250 compatible devices clear the 'UART_LST_TEMT" bit together with the "UART_LSR_THRE" bit when writing data to the device.

The NetMos device does it slightly different

I believe that the TEMT bit is coupled to the shift register. The problem is that after writing data to the device and very quickly after that one does call serial8250_tx_empty, it returns the wrong information.

My patch makes the test more robust (and solves the problem) and it does not affect the already correct devices.

Alan:

We may yet need to quirk this but now we know which chips we have a way to do that should we find this breaks some other 8250 clone with dodgy THRE.

---
 drivers/serial/8250.c |    7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
drivers/serial/8250.c +4 -3
@@ -83,6 +83,9 @@ static unsigned int skip_txen_test; /* f
 
 #define PASS_LIMIT	256
 
+#define BOTH_EMPTY 	(UART_LSR_TEMT | UART_LSR_THRE)
+
+
 /*
  * We default to IRQ0 for the "no irq" hack.   Some
  * machine types want others as well - they're free
@@ -1792,7 +1795,7 @@ static unsigned int serial8250_tx_empty(
 	up->lsr_saved_flags |= lsr & LSR_SAVE_FLAGS;
 	spin_unlock_irqrestore(&up->port.lock, flags);
 
-	return lsr & UART_LSR_TEMT ? TIOCSER_TEMT : 0;
+	return (lsr & BOTH_EMPTY) == BOTH_EMPTY ? TIOCSER_TEMT : 0;
 }
 
 static unsigned int serial8250_get_mctrl(struct uart_port *port)
@@ -1850,8 +1853,6 @@ static void serial8250_break_ctl(struct
 	spin_unlock_irqrestore(&up->port.lock, flags);
 }
 
-#define BOTH_EMPTY (UART_LSR_TEMT | UART_LSR_THRE)
-
 /*
  *	Wait for transmitter & holding register to empty
  */
author
#80 lore raw

PATCH 77/93 sysfs: sysfs_sd_setattr set iattrs unconditionally

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Eric W. Biederman <[email protected]>

commit 7c0ff870d1ed287504a61ed865f3d728c757436b upstream.

There is currently a bug in sysfs_sd_setattr inherited from sysfs_setattr in 2.6.32 where the first time we set the attributes on a sysfs file we allocate backing store but do not set the backing store attributes. Resulting in overly restrictive permissions on sysfs files.

The fix is to simply modify the code so that it always executes when we update the sysfs attributes, as we did in 2.6.31 and earlier.

---
 fs/sysfs/inode.c |   43 +++++++++++++++++++++----------------------
 1 file changed, 21 insertions(+), 22 deletions(-)
fs/sysfs/inode.c +21 -22
@@ -94,30 +94,29 @@ int sysfs_setattr(struct dentry * dentry
 		if (!sd_attrs)
 			return -ENOMEM;
 		sd->s_iattr = sd_attrs;
-	} else {
-		/* attributes were changed at least once in past */
-		iattrs = &sd_attrs->ia_iattr;
+	}
+	/* attributes were changed at least once in past */
+	iattrs = &sd_attrs->ia_iattr;
 
-		if (ia_valid & ATTR_UID)
-			iattrs->ia_uid = iattr->ia_uid;
-		if (ia_valid & ATTR_GID)
-			iattrs->ia_gid = iattr->ia_gid;
-		if (ia_valid & ATTR_ATIME)
-			iattrs->ia_atime = timespec_trunc(iattr->ia_atime,
-					inode->i_sb->s_time_gran);
-		if (ia_valid & ATTR_MTIME)
-			iattrs->ia_mtime = timespec_trunc(iattr->ia_mtime,
-					inode->i_sb->s_time_gran);
-		if (ia_valid & ATTR_CTIME)
-			iattrs->ia_ctime = timespec_trunc(iattr->ia_ctime,
-					inode->i_sb->s_time_gran);
-		if (ia_valid & ATTR_MODE) {
-			umode_t mode = iattr->ia_mode;
+	if (ia_valid & ATTR_UID)
+		iattrs->ia_uid = iattr->ia_uid;
+	if (ia_valid & ATTR_GID)
+		iattrs->ia_gid = iattr->ia_gid;
+	if (ia_valid & ATTR_ATIME)
+		iattrs->ia_atime = timespec_trunc(iattr->ia_atime,
+			inode->i_sb->s_time_gran);
+	if (ia_valid & ATTR_MTIME)
+		iattrs->ia_mtime = timespec_trunc(iattr->ia_mtime,
+			inode->i_sb->s_time_gran);
+	if (ia_valid & ATTR_CTIME)
+		iattrs->ia_ctime = timespec_trunc(iattr->ia_ctime,
+			inode->i_sb->s_time_gran);
+	if (ia_valid & ATTR_MODE) {
+		umode_t mode = iattr->ia_mode;
 
-			if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
-				mode &= ~S_ISGID;
-			iattrs->ia_mode = sd->s_mode = mode;
-		}
+		if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
+			mode &= ~S_ISGID;
+		iattrs->ia_mode = sd->s_mode = mode;
 	}
 	return error;
 }
author
#81 lore raw

PATCH 78/93 class: Free the class private data in class_release

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Laurent Pinchart <[email protected]>

commit 18d19c96457d172d913510c083bc7411ed40cb10 upstream.

Fix a memory leak by freeing the memory allocated in __class_register for the class private data.

---
 drivers/base/class.c |    2 ++
 1 file changed, 2 insertions(+)
drivers/base/class.c +2 -0
@@ -59,6 +59,8 @@ static void class_release(struct kobject
 	else
 		pr_debug("class '%s' does not have a release() function, "
 			 "be careful\n", class->name);
+
+	kfree(cp);
 }
 
 static struct sysfs_ops class_sysfs_ops = {
author
#82 lore raw

PATCH 79/93 USB: usbfs: only copy the actual data received

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Greg KH <[email protected]>

commit d4a4683ca054ed9917dfc9e3ff0f7ecf74ad90d6 upstream.

We need to only copy the data received by the device to userspace, not the whole kernel buffer, which can contain "stale" data.

Thanks to Marcus Meissner for pointing this out and testing the fix.

---
 drivers/usb/core/devio.c |    8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/usb/core/devio.c +4 -4
@@ -1312,9 +1312,9 @@ static int processcompl(struct async *as
 	void __user *addr = as->userurb;
 	unsigned int i;
 
-	if (as->userbuffer)
+	if (as->userbuffer && urb->actual_length)
 		if (copy_to_user(as->userbuffer, urb->transfer_buffer,
-				 urb->transfer_buffer_length))
+				 urb->actual_length))
 			goto err_out;
 	if (put_user(as->status, &userurb->status))
 		goto err_out;
@@ -1435,9 +1435,9 @@ static int processcompl_compat(struct as
 	void __user *addr = as->userurb;
 	unsigned int i;
 
-	if (as->userbuffer)
+	if (as->userbuffer && urb->actual_length)
 		if (copy_to_user(as->userbuffer, urb->transfer_buffer,
-				 urb->transfer_buffer_length))
+				 urb->actual_length))
 			return -EFAULT;
 	if (put_user(as->status, &userurb->status))
 		return -EFAULT;
author
#83 lore raw

PATCH 80/93 USB: usbfs: properly clean up the as structure on error paths

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Linus Torvalds <[email protected]>

commit ddeee0b2eec2a51b0712b04de4b39e7bec892a53 upstream.

I notice that the processcompl_compat() function seems to be leaking the 'struct async *as' in the error paths.

I think that the calling convention is fundamentally buggered. The caller is the one that did the "reap_as()" to get the as thing, the caller should be the one to free it too.

Freeing it in the caller also means that it very clearly always gets freed, and avoids the need for any "free in the error case too".

From: Linus Torvalds <[email protected]>

Cc Alan Stern <[email protected]>
Cc Marcus Meissner <[email protected]>
---
 drivers/usb/core/devio.c |   40 ++++++++++++++++++++++++++--------------
 1 file changed, 26 insertions(+), 14 deletions(-)
drivers/usb/core/devio.c +26 -14
@@ -1334,14 +1334,11 @@ static int processcompl(struct async *as
 		}
 	}
 
-	free_async(as);
-
 	if (put_user(addr, (void __user * __user *)arg))
 		return -EFAULT;
 	return 0;
 
 err_out:
-	free_async(as);
 	return -EFAULT;
 }
 
@@ -1371,8 +1368,11 @@ static struct async *reap_as(struct dev_
 static int proc_reapurb(struct dev_state *ps, void __user *arg)
 {
 	struct async *as = reap_as(ps);
-	if (as)
-		return processcompl(as, (void __user * __user *)arg);
+	if (as) {
+		int retval = processcompl(as, (void __user * __user *)arg);
+		free_async(as);
+		return retval;
+	}
 	if (signal_pending(current))
 		return -EINTR;
 	return -EIO;
@@ -1380,11 +1380,16 @@ static int proc_reapurb(struct dev_state
 
 static int proc_reapurbnonblock(struct dev_state *ps, void __user *arg)
 {
+	int retval;
 	struct async *as;
 
-	if (!(as = async_getcompleted(ps)))
-		return -EAGAIN;
-	return processcompl(as, (void __user * __user *)arg);
+	as = async_getcompleted(ps);
+	retval = -EAGAIN;
+	if (as) {
+		retval = processcompl(as, (void __user * __user *)arg);
+		free_async(as);
+	}
+	return retval;
 }
 
 #ifdef CONFIG_COMPAT
@@ -1457,7 +1462,6 @@ static int processcompl_compat(struct as
 		}
 	}
 
-	free_async(as);
 	if (put_user(ptr_to_compat(addr), (u32 __user *)arg))
 		return -EFAULT;
 	return 0;
@@ -1466,8 +1470,11 @@ static int processcompl_compat(struct as
 static int proc_reapurb_compat(struct dev_state *ps, void __user *arg)
 {
 	struct async *as = reap_as(ps);
-	if (as)
-		return processcompl_compat(as, (void __user * __user *)arg);
+	if (as) {
+		int retval = processcompl_compat(as, (void __user * __user *)arg);
+		free_async(as);
+		return retval;
+	}
 	if (signal_pending(current))
 		return -EINTR;
 	return -EIO;
@@ -1475,11 +1482,16 @@ static int proc_reapurb_compat(struct de
 
 static int proc_reapurbnonblock_compat(struct dev_state *ps, void __user *arg)
 {
+	int retval;
 	struct async *as;
 
-	if (!(as = async_getcompleted(ps)))
-		return -EAGAIN;
-	return processcompl_compat(as, (void __user * __user *)arg);
+	retval = -EAGAIN;
+	as = async_getcompleted(ps);
+	if (as) {
+		retval = processcompl_compat(as, (void __user * __user *)arg);
+		free_async(as);
+	}
+	return retval;
 }
 
 #endif
author
#84 lore raw

PATCH 81/93 rtl8187: Add new device ID

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Jelle Martijn Kok <[email protected]>

commit 174b24963eaf96dc5e093502ee09639aed13eb2f upstream.

Add new RTL8187B device.

---
 drivers/net/wireless/rtl818x/rtl8187_dev.c |    1 +
 1 file changed, 1 insertion(+)
drivers/net/wireless/rtl818x/rtl8187_dev.c +1 -0
@@ -65,6 +65,7 @@ static struct usb_device_id rtl8187_tabl
 	/* Sitecom */
 	{USB_DEVICE(0x0df6, 0x000d), .driver_info = DEVICE_RTL8187},
 	{USB_DEVICE(0x0df6, 0x0028), .driver_info = DEVICE_RTL8187B},
+	{USB_DEVICE(0x0df6, 0x0029), .driver_info = DEVICE_RTL8187B},
 	/* Sphairon Access Systems GmbH */
 	{USB_DEVICE(0x114B, 0x0150), .driver_info = DEVICE_RTL8187},
 	/* Dick Smith Electronics */
author
#85 lore raw

PATCH 82/93 ACPI: Add NULL pointer check in acpi_bus_start

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Thomas Renninger <[email protected]>

commit d2f6650a950dadd20667a04a9dc785f240d43695 upstream.

If acpi_bus_add does not return a device and it's passed to acpi_bus_start, bad things will happen:

BUG: unable to handle kernel NULL pointer dereference at 0000000000000008
IP: [<ffffffff8128402d>] acpi_bus_start+0x14/0x24
...
[<ffffffffa008977a>] acpiphp_bus_add+0xba/0x130 [acpiphp]
[<ffffffffa008aa72>] enable_device+0x132/0x2ff [acpiphp]
[<ffffffffa0089b68>] acpiphp_enable_slot+0xb8/0x130 [acpiphp]
[<ffffffffa0089df7>] handle_hotplug_event_func+0x87/0x190 [acpiphp]

Next patch would make this NULL pointer check obsolete, but better having one more than one missing...

---
 drivers/acpi/scan.c |    3 +++
 1 file changed, 3 insertions(+)
drivers/acpi/scan.c +3 -0
@@ -1357,6 +1357,9 @@ int acpi_bus_start(struct acpi_device *d
 {
 	struct acpi_bus_ops ops;
 
+	if (!device)
+		return -EINVAL;
+
 	memset(&ops, 0, sizeof(ops));
 	ops.acpi_op_start = 1;
 
author
#86 lore raw

PATCH 83/93 ACPI: fix High cpu temperature with 2.6.32

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Arjan van de Ven <[email protected]>

commit 370d5cd88509b93b76eb2f5f97efbd71c25061cb upstream.

Since the rewrite of the CPU idle governor in 2.6.32, two laptops have surfaced where the BIOS advertises a C2 power state, but for some reason this state is not functioning (as verified in both cases by powertop before the patch in .32).

The old governor had the accidental behavior that if a non-working state was chosen too many times, it would end up falling back to C1. The new governor works differently and this accidental behavior is no longer there; the result is a high temperature on these two machines.

This patch adds these 2 machines to the DMI table for C state anomalies; by just not using C2 both these machines are better off (the TSC can be used instead of the pm timer, giving a performance boost for example).

Addresses http://bugzilla.kernel.org/show_bug.cgi?id=14742

---
 drivers/acpi/processor_idle.c |    8 ++++++++
 1 file changed, 8 insertions(+)
drivers/acpi/processor_idle.c +8 -0
@@ -110,6 +110,14 @@ static struct dmi_system_id __cpuinitdat
 	  DMI_MATCH(DMI_BIOS_VENDOR,"Phoenix Technologies LTD"),
 	  DMI_MATCH(DMI_BIOS_VERSION,"SHE845M0.86C.0013.D.0302131307")},
 	 (void *)2},
+	{ set_max_cstate, "Pavilion zv5000", {
+	  DMI_MATCH(DMI_SYS_VENDOR, "Hewlett-Packard"),
+	  DMI_MATCH(DMI_PRODUCT_NAME,"Pavilion zv5000 (DS502A#ABA)")},
+	 (void *)1},
+	{ set_max_cstate, "Asus L8400B", {
+	  DMI_MATCH(DMI_SYS_VENDOR, "ASUSTeK Computer Inc."),
+	  DMI_MATCH(DMI_PRODUCT_NAME,"L8400B series Notebook PC")},
+	 (void *)1},
 	{},
 };
 
author
#87 lore raw

PATCH 84/93 drm/radeon/kms: use udelay for short delays

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Dave Airlie <[email protected]>

commit 01d4503968f471f876fb44335800d2cf8dc5a2ce upstream.

For usec delays use udelay instead of scheduling, this should allow reclocking to happen faster. This also was the cause of reported 33s delays at bootup on certain systems.

Fixes freedesktop.org bug 25506
---
 drivers/gpu/drm/radeon/atom.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/radeon/atom.c +1 -1
@@ -607,7 +607,7 @@ static void atom_op_delay(atom_exec_cont
 	uint8_t count = U8((*ptr)++);
 	SDEBUG("   count: %d\n", count);
 	if (arg == ATOM_UNIT_MICROSEC)
-		schedule_timeout_uninterruptible(usecs_to_jiffies(count));
+		udelay(count);
 	else
 		schedule_timeout_uninterruptible(msecs_to_jiffies(count));
 }
author
#88 lore raw

PATCH 85/93 NFS: Too many GETATTR and ACCESS calls after direct I/O

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Chuck Lever <[email protected]>

commit 65d269538a1129495ac45a14a777cd11cfe881d8 upstream.

The cached read and write paths initialize fattr->time_start in their setup procedures. The value of fattr->time_start is propagated to read_cache_jiffies by nfs_update_inode(). Subsequent calls to nfs_attribute_timeout() will then use a good time stamp when computing the attribute cache timeout, and squelch unneeded GETATTR calls.

Since the direct I/O paths erroneously leave the inode's fattr->time_start field set to zero, read_cache_jiffies for that inode is set to zero after any direct read or write operation. This triggers an otw GETATTR or ACCESS call to update the file's attribute and access caches properly, even when the NFS READ or WRITE replies have usable post-op attributes.

Make sure the direct read and write setup code performs the same fattr initialization as the cached I/O paths to prevent unnecessary GETATTR calls.

This was likely introduced by commit 0e574af1 in 2.6.15, which appears to add new nfs_fattr_init() call sites in the cached read and write paths, but not in the equivalent places in fs/nfs/direct.c. A subsequent commit in the same series, 33801147, introduces the fattr->time_start field.

Interestingly, the direct write reschedule path already has a call to nfs_fattr_init() in the right place.

---
 fs/nfs/direct.c |    3 +++
 1 file changed, 3 insertions(+)
fs/nfs/direct.c +3 -0
@@ -342,6 +342,7 @@ static ssize_t nfs_direct_read_schedule_
 		data->res.fattr = &data->fattr;
 		data->res.eof = 0;
 		data->res.count = bytes;
+		nfs_fattr_init(&data->fattr);
 		msg.rpc_argp = &data->args;
 		msg.rpc_resp = &data->res;
 
@@ -575,6 +576,7 @@ static void nfs_direct_commit_schedule(s
 	data->res.count = 0;
 	data->res.fattr = &data->fattr;
 	data->res.verf = &data->verf;
+	nfs_fattr_init(&data->fattr);
 
 	NFS_PROTO(data->inode)->commit_setup(data, &msg);
 
@@ -766,6 +768,7 @@ static ssize_t nfs_direct_write_schedule
 		data->res.fattr = &data->fattr;
 		data->res.count = bytes;
 		data->res.verf = &data->verf;
+		nfs_fattr_init(&data->fattr);
 
 		task_setup_data.task = &data->task;
 		task_setup_data.callback_data = data;
author
#89 lore raw

PATCH 86/93 eCryptfs: Add getattr function

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Tyler Hicks <[email protected]>

commit f8f484d1b6677dd5cd5e7e605db747e8c30bbd47 upstream.

The i_blocks field of an eCryptfs inode cannot be trusted, but generic_fillattr() uses it to instantiate the blocks field of a stat() syscall when a filesystem doesn't implement its own getattr(). Users have noticed that the output of du is incorrect on newly created files.

This patch creates ecryptfs_getattr() which calls into the lower filesystem's getattr() so that eCryptfs can use its kstat.blocks value after calling generic_fillattr(). It is important to note that the block count includes the eCryptfs metadata stored in the beginning of the lower file plus any padding used to fill an extent before encryption.

https://bugs.launchpad.net/ecryptfs/+bug/390833

---
 fs/ecryptfs/inode.c |   16 ++++++++++++++++
 1 file changed, 16 insertions(+)
fs/ecryptfs/inode.c +16 -0
@@ -971,6 +971,21 @@ out:
 	return rc;
 }
 
+int ecryptfs_getattr(struct vfsmount *mnt, struct dentry *dentry,
+		     struct kstat *stat)
+{
+	struct kstat lower_stat;
+	int rc;
+
+	rc = vfs_getattr(ecryptfs_dentry_to_lower_mnt(dentry),
+			 ecryptfs_dentry_to_lower(dentry), &lower_stat);
+	if (!rc) {
+		generic_fillattr(dentry->d_inode, stat);
+		stat->blocks = lower_stat.blocks;
+	}
+	return rc;
+}
+
 int
 ecryptfs_setxattr(struct dentry *dentry, const char *name, const void *value,
 		  size_t size, int flags)
@@ -1100,6 +1115,7 @@ const struct inode_operations ecryptfs_d
 const struct inode_operations ecryptfs_main_iops = {
 	.permission = ecryptfs_permission,
 	.setattr = ecryptfs_setattr,
+	.getattr = ecryptfs_getattr,
 	.setxattr = ecryptfs_setxattr,
 	.getxattr = ecryptfs_getxattr,
 	.listxattr = ecryptfs_listxattr,
author
#90 lore raw

PATCH 87/93 b43: Fix throughput regression

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Larry Finger <[email protected]>

commit b6c3f5be7c6ac3375f44de4545c1ffe216b34022 upstream.

Commit c7ab5ef9bcd281135c21b4732c9be779585181be entitled "b43: implement short slot and basic rate handling" reduced the transmit throughput for my BCM4311 device from 18 Mb/s to 0.7 Mb/s. The basic rate handling portion is OK, the problem is in the short slot handling.

Prior to this change, the short slot enable/disable routines were never called. Experimentation showed that the critical part was changing the value at offset 0x0010 in the shared memory. This is supposed to contain the 802.11 Slot Time in usec, but if it is changed from its initial value of zero, performance is destroyed. On the other hand, changing the value in the MMIO register corresponding to the Interframe Slot Time increased performance from 18 to 22 Mb/s. A BCM4306/3 also shows dramatic improvement of the transmit rate from 5.3 to 19.0 Mb/s.

Other changes in the patch include removal of the magic number for the MMIO register, and allowing the slot time to be set for any PHY operating in the 2.4 GHz band. Previously, the routine was executed only for G PHYs.

---
 drivers/net/wireless/b43/b43.h  |    1 +
 drivers/net/wireless/b43/main.c |   13 ++++++++++---
 2 files changed, 11 insertions(+), 3 deletions(-)
drivers/net/wireless/b43/b43.h +1 -0
@@ -117,6 +117,7 @@
 #define B43_MMIO_TSF_2			0x636	/* core rev < 3 only */
 #define B43_MMIO_TSF_3			0x638	/* core rev < 3 only */
 #define B43_MMIO_RNG			0x65A
+#define B43_MMIO_IFSSLOT		0x684	/* Interframe slot time */
 #define B43_MMIO_IFSCTL			0x688 /* Interframe space control */
 #define  B43_MMIO_IFSCTL_USE_EDCF	0x0004
 #define B43_MMIO_POWERUP_DELAY		0x6A8
drivers/net/wireless/b43/main.c +10 -3
@@ -628,10 +628,17 @@ static void b43_upload_card_macaddress(s
 static void b43_set_slot_time(struct b43_wldev *dev, u16 slot_time)
 {
 	/* slot_time is in usec. */
-	if (dev->phy.type != B43_PHYTYPE_G)
+	/* This test used to exit for all but a G PHY. */
+	if (b43_current_band(dev->wl) == IEEE80211_BAND_5GHZ)
 		return;
-	b43_write16(dev, 0x684, 510 + slot_time);
-	b43_shm_write16(dev, B43_SHM_SHARED, 0x0010, slot_time);
+	b43_write16(dev, B43_MMIO_IFSSLOT, 510 + slot_time);
+	/* Shared memory location 0x0010 is the slot time and should be
+	 * set to slot_time; however, this register is initially 0 and changing
+	 * the value adversely affects the transmit rate for BCM4311
+	 * devices. Until this behavior is unterstood, delete this step
+	 *
+	 * b43_shm_write16(dev, B43_SHM_SHARED, 0x0010, slot_time);
+	 */
 }
 
 static void b43_short_slot_timing_enable(struct b43_wldev *dev)
author
#91 lore raw

PATCH 88/93 ath9k: Fix sequence numbers for PAE frames

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Sujith <[email protected]>

commit 6c8afef551fef87a3bf24f8a74c69a7f2f72fc82 upstream.

Currently, PAE frames are not assigned proper sequence numbers. Since sending PAE frames as part of aggregates breaks crupto with several APs, they are sent as normal MPDUs. Fix the seqeuence number issue by updating the frame with the internal sequence number.

---
 drivers/net/wireless/ath/ath9k/xmit.c |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/net/wireless/ath/ath9k/xmit.c +2 -2
@@ -1563,7 +1563,7 @@ static int ath_tx_setup_buffer(struct ie
 
 	bf->bf_frmlen = skb->len + FCS_LEN - (hdrlen & 3);
 
-	if (conf_is_ht(&sc->hw->conf) && !is_pae(skb))
+	if (conf_is_ht(&sc->hw->conf))
 		bf->bf_state.bf_type |= BUF_HT;
 
 	bf->bf_flags = setup_tx_flags(sc, skb, txctl->txq);
@@ -1648,7 +1648,7 @@ static void ath_tx_start_dma(struct ath_
 			goto tx_done;
 		}
 
-		if (tx_info->flags & IEEE80211_TX_CTL_AMPDU) {
+		if ((tx_info->flags & IEEE80211_TX_CTL_AMPDU) && !is_pae(skb)) {
 			/*
 			 * Try aggregation if it's a unicast data frame
 			 * and the destination is HT capable.
author
#92 lore raw

PATCH 89/93 mac80211: Fix probe request filtering in IBSS mode

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Benoit Papillault <[email protected]>

commit 0da780c269957783d341fc3559e6b4c9912af7b4 upstream.

We only reply to probe request if either the requested SSID is the broadcast SSID or if the requested SSID matches our own SSID. This latter case was not properly handled since we were replying to different SSID with the same length as our own SSID.

---
 net/mac80211/ibss.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
net/mac80211/ibss.c +1 -1
@@ -643,7 +643,7 @@ static void ieee80211_rx_mgmt_probe_req(
 	}
 	if (pos[1] != 0 &&
 	    (pos[1] != ifibss->ssid_len ||
-	     !memcmp(pos + 2, ifibss->ssid, ifibss->ssid_len))) {
+	     memcmp(pos + 2, ifibss->ssid, ifibss->ssid_len))) {
 		/* Ignore ProbeReq for foreign SSID */
 		return;
 	}
author
#93 lore raw

PATCH 90/93 iwlwifi: Fix to set correct ht configuration

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Wey-Yi Guy <[email protected]>

commit 098dfded5b1b09927995e89c6d689f85a0f53384 upstream.

iwl_set_rxon_ht() only get called in iwl_post_associate(); which cause possible incorrect ht configuration. Adding the call in iwl_mac_config() if IEEE80211_CONF_CHANGE_CHANNEL flag is set to re-configure and send rxon command.

Fixes
http://bugzilla.intellinuxwireless.org/show_bug.cgi?id=2146

---
 drivers/net/wireless/iwlwifi/iwl-core.c |    1 +
 1 file changed, 1 insertion(+)
drivers/net/wireless/iwlwifi/iwl-core.c +1 -0
@@ -2645,6 +2645,7 @@ int iwl_mac_config(struct ieee80211_hw *
 		if ((le16_to_cpu(priv->staging_rxon.channel) != ch))
 			priv->staging_rxon.flags = 0;
 
+		iwl_set_rxon_ht(priv, ht_conf);
 		iwl_set_rxon_channel(priv, conf->channel);
 
 		iwl_set_flags_for_band(priv, conf->channel->band);
author
#94 lore raw

PATCH 91/93 dm stripe: avoid divide by zero with invalid stripe count

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Nikanth Karthikesan <[email protected]>

commit 781248c1b50c776a9ef4be1130f84ced1cba42fe upstream.

If a table containing zero as stripe count is passed into stripe_ctr the code attempts to divide by zero.

This patch changes DM_TABLE_LOAD to return -EINVAL if the stripe count is zero.

We now get the following error messages:
device-mapper: table: 253:0: striped: Invalid stripe count
device-mapper: ioctl: error adding target to table

---
 drivers/md/dm-stripe.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/md/dm-stripe.c +1 -1
@@ -110,7 +110,7 @@ static int stripe_ctr(struct dm_target *
 	}
 
 	stripes = simple_strtoul(argv[0], &end, 10);
-	if (*end) {
+	if (!stripes || *end) {
 		ti->error = "Invalid stripe count";
 		return -EINVAL;
 	}
author
#95 lore raw

PATCH 92/93 dm log: userspace fix overhead_size calcuations

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Jonathan Brassow <[email protected]>

commit ebfd32bba9b518d684009d9d21a56742337ca1b3 upstream.

This patch fixes two bugs that revolve around the miscalculation and misuse of the variable 'overhead_size'. 'overhead_size' is the size of the various header structures used during communication.

The first bug is the use of 'sizeof' with the pointer of a structure instead of the structure itself - resulting in the wrong size being computed. This is then used in a check to see if the payload (data_size) would be to large for the preallocated structure. Since the bug produces a smaller value for the overhead, it was possible for the structure to be breached. (Although the current users of the code do not currently send enough data to trigger this bug.)

The second bug is that the 'overhead_size' value is used to compute how much of the preallocated space should be cleared before populating it with fresh data. This should have simply been 'sizeof(struct cn_msg)' not overhead_size. The fact that 'overhead_size' was computed incorrectly made this problem "less bad" - leaving only a pointer's worth of space at the end uncleared. Thus, this bug was never producing a bad result, but still needs to be fixed - especially now that the value is computed correctly.

---
 drivers/md/dm-log-userspace-transfer.c |   10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)
drivers/md/dm-log-userspace-transfer.c +7 -3
@@ -172,11 +172,15 @@ int dm_consult_userspace(const char *uui
 {
 	int r = 0;
 	size_t dummy = 0;
-	int overhead_size =
-		sizeof(struct dm_ulog_request *) + sizeof(struct cn_msg);
+	int overhead_size = sizeof(struct dm_ulog_request) + sizeof(struct cn_msg);
 	struct dm_ulog_request *tfr = prealloced_ulog_tfr;
 	struct receiving_pkg pkg;
 
+	/*
+	 * Given the space needed to hold the 'struct cn_msg' and
+	 * 'struct dm_ulog_request' - do we have enough payload
+	 * space remaining?
+	 */
 	if (data_size > (DM_ULOG_PREALLOCED_SIZE - overhead_size)) {
 		DMINFO("Size of tfr exceeds preallocated size");
 		return -EINVAL;
@@ -191,7 +195,7 @@ resend:
 	 */
 	mutex_lock(&dm_ulog_lock);
 
-	memset(tfr, 0, DM_ULOG_PREALLOCED_SIZE - overhead_size);
+	memset(tfr, 0, DM_ULOG_PREALLOCED_SIZE - sizeof(struct cn_msg));
 	memcpy(tfr->uuid, uuid, DM_UUID_LEN);
 	tfr->luid = luid;
 	tfr->seq = dm_ulog_seq++;
author
#96 lore raw

PATCH 93/93 dm mpath: fix stall when requeueing io

In reply to Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Kiyoshi Ueda <[email protected]>

commit 9eef87da2a8ea4920e0d913ff977cac064b68ee0 upstream.

This patch fixes the problem that system may stall if target's ->map_rq returns DM_MAPIO_REQUEUE in map_request().
E.g. stall happens on 1 CPU box when a dm-mpath device with queue_if_no_path
bounces between all-paths-down and paths-up on I/O load.

When target's ->map_rq returns DM_MAPIO_REQUEUE, map_request() requeues the request and returns to dm_request_fn(). Then, dm_request_fn() doesn't exit the I/O dispatching loop and continues processing the requeued request again.
This map and requeue loop can be done with interrupt disabled, so 1 CPU system can be stalled if this situation happens.

For example, commands below can stall my 1 CPU box within 1 minute or so:

  # dmsetup table mp
  mp: 0 2097152 multipath 1 queue_if_no_path 0 1 1 service-time 0 1 2 8:144 1 1
  # while true; do dd if=/dev/mapper/mp of=/dev/null bs=1M count=100; done &
  # while true; do \
Quoted 5 lines

dmsetup message mp 0 "fail_path 8:144" \
dmsetup suspend --noflush mp \
dmsetup resume mp \
dmsetup message mp 0 "reinstate_path 8:144" \
done

To fix the problem above, this patch changes dm_request_fn() to exit the I/O dispatching loop once if a request is requeued in map_request().

---
 drivers/md/dm.c |   21 ++++++++++++++++-----
 1 file changed, 16 insertions(+), 5 deletions(-)
drivers/md/dm.c +16 -5
@@ -1487,11 +1487,15 @@ static int dm_prep_fn(struct request_que
 	return BLKPREP_OK;
 }
 
-static void map_request(struct dm_target *ti, struct request *rq,
-			struct mapped_device *md)
+/*
+ * Returns:
+ * 0  : the request has been processed (not requeued)
+ * !0 : the request has been requeued
+ */
+static int map_request(struct dm_target *ti, struct request *clone,
+		       struct mapped_device *md)
 {
-	int r;
-	struct request *clone = rq->special;
+	int r, requeued = 0;
 	struct dm_rq_target_io *tio = clone->end_io_data;
 
 	/*
@@ -1516,6 +1520,7 @@ static void map_request(struct dm_target
 	case DM_MAPIO_REQUEUE:
 		/* The target wants to requeue the I/O */
 		dm_requeue_unmapped_request(clone);
+		requeued = 1;
 		break;
 	default:
 		if (r > 0) {
@@ -1527,6 +1532,8 @@ static void map_request(struct dm_target
 		dm_kill_unmapped_request(clone, r);
 		break;
 	}
+
+	return requeued;
 }
 
 /*
@@ -1568,12 +1575,16 @@ static void dm_request_fn(struct request
 
 		blk_start_request(rq);
 		spin_unlock(q->queue_lock);
-		map_request(ti, rq, md);
+		if (map_request(ti, rq, md))
+			goto requeued;
 		spin_lock_irq(q->queue_lock);
 	}
 
 	goto out;
 
+requeued:
+	spin_lock_irq(q->queue_lock);
+
 plug_and_out:
 	if (!elv_queue_empty(q))
 		/* Some requests still remain, retry later */
#97 lore raw

Re: [Stable-review] [93/93] dm mpath: fix stall when requeueing io

Quoted 51 lines · Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Kiyoshi Ueda <[email protected]>

commit 9eef87da2a8ea4920e0d913ff977cac064b68ee0 upstream.

This patch fixes the problem that system may stall if target's ->map_rq returns DM_MAPIO_REQUEUE in map_request().
E.g. stall happens on 1 CPU box when a dm-mpath device with queue_if_no_path
bounces between all-paths-down and paths-up on I/O load.

When target's ->map_rq returns DM_MAPIO_REQUEUE, map_request() requeues the request and returns to dm_request_fn(). Then, dm_request_fn() doesn't exit the I/O dispatching loop and continues processing the requeued request again.
This map and requeue loop can be done with interrupt disabled, so 1 CPU system can be stalled if this situation happens.

For example, commands below can stall my 1 CPU box within 1 minute or so:

  # dmsetup table mp
  mp: 0 2097152 multipath 1 queue_if_no_path 0 1 1 service-time 0 1 2 8:144 1 1
  # while true; do dd if=/dev/mapper/mp of=/dev/null bs=1M count=100; done &
  # while true; do \
Quoted 5 lines

dmsetup message mp 0 "fail_path 8:144" \
dmsetup suspend --noflush mp \
dmsetup resume mp \
dmsetup message mp 0 "reinstate_path 8:144" \
done

To fix the problem above, this patch changes dm_request_fn() to exit the I/O dispatching loop once if a request is requeued in map_request().

---
 drivers/md/dm.c |   21 ++++++++++++++++-----
 1 file changed, 16 insertions(+), 5 deletions(-)
drivers/md/dm.c +0 -2
@@ -1487,11 +1487,15 @@ static int dm_prep_fn(struct request_que
 	return BLKPREP_OK;
 }
 
-static void map_request(struct dm_target *ti, struct request *rq,
-			struct mapped_device *md)

drivers/md/dm.c static int dm_prep_fn(struct request_que · line 1494

Quoted 10 lines
drivers/md/dm.c +7 -2
+/*
+ * Returns:
+ * 0  : the request has been processed (not requeued)
+ * !0 : the request has been requeued
+ */
+static int map_request(struct dm_target *ti, struct request *clone,
+		       struct mapped_device *md)
 {
-	int r;
-	struct request *clone = rq->special;

This change requires the argument to this function to be a rq->special pointer. This is changed in the map_request function by the following patch:

commit b4324feeae304ae39e631a254d238a7d63be004b

Author: Kiyoshi Ueda <[email protected]>
Date:   Thu Dec 10 23:52:16 2009 +0000

dm: use md pending for in flight IO counting

Quoted 22 lines
drivers/md/dm.c +4 -0
+	int r, requeued = 0;
 	struct dm_rq_target_io *tio = clone->end_io_data;
 
 	/*
@@ -1516,6 +1520,7 @@ static void map_request(struct dm_target
 	case DM_MAPIO_REQUEUE:
 		/* The target wants to requeue the I/O */
 		dm_requeue_unmapped_request(clone);
+		requeued = 1;
 		break;
 	default:
 		if (r > 0) {
@@ -1527,6 +1532,8 @@ static void map_request(struct dm_target
 		dm_kill_unmapped_request(clone, r);
 		break;
 	}
+
+	return requeued;
 }
 
 /*

drivers/md/dm.c static void dm_request_fn(struct request · line 1581

Quoted 8 lines
drivers/md/dm.c +2 -1
@@ -1568,12 +1575,16 @@ static void dm_request_fn(struct request
 
 		blk_start_request(rq);
 		spin_unlock(q->queue_lock);
-		map_request(ti, rq, md);
+		if (map_request(ti, rq, md))
+			goto requeued;
 		spin_lock_irq(q->queue_lock);
 	}

That is the current state of dm_request_function:

                clone = rq->special;
                atomic_inc(&md->pending[rq_data_dir(clone)]);

                spin_unlock(q->queue_lock);
                if (map_request(ti, clone, md))

While looking over the code I also noticed that the spinlock is dropped with spin_unlock and then reacquired with spin_lock_irq. Isn't the irq version too much in that case? Or was the intention to have interrupts enabled when unlocking?

-Stefan

Quoted 14 lines
drivers/md/dm.c +3 -0
 	goto out;
 
+requeued:
+	spin_lock_irq(q->queue_lock);
+
 plug_and_out:
 	if (!elv_queue_empty(q))
 		/* Some requests still remain, retry later */
 

_______________________________________________
Stable-review mailing list
[email protected]
http://linux.kernel.org/mailman/listinfo/stable-review

#98 lore raw

Re: [Stable-review] [93/93] dm mpath: fix stall when requeueing io

Hi Alasdair, Greg,

Please replace this patch with the patch attached below. This patch seems to have been broken somewhere, since the original patch (*) for 2.6.33-rc6 cosmetically depends on another patch included in 2.6.33-rc1.
(*) http://marc.info/?l=dm-devel&m=126518144727377&w=2

Stefan, thank you for spotting this.

BTW, as for your spin lock question, the spin_lock_irq could be spin_lock. Please see my answer below for details.

Quoted 18 lines · Stefan Bader

Review static void dm_request_fn(struct request · line 1581

Quoted 9 lines
(quoted diff) +2 -1
@@ -1568,12 +1575,16 @@ static void dm_request_fn(struct request
 
 		blk_start_request(rq);
 		spin_unlock(q->queue_lock);
-		map_request(ti, rq, md);
+		if (map_request(ti, rq, md))
+			goto requeued;
 		spin_lock_irq(q->queue_lock);
 	}

That is the current state of dm_request_function:

                clone = rq->special;
                atomic_inc(&md->pending[rq_data_dir(clone)]);

                spin_unlock(q->queue_lock);
                if (map_request(ti, clone, md))
Quoted 3 lines

While looking over the code I also noticed that the spinlock is dropped with spin_unlock and then reacquired with spin_lock_irq. Isn't the irq version too much in that case? Or was the intention to have interrupts enabled when unlocking?

In the current device-mapper code, I would like to go with spin_unlock/lock here.
However, there was a case to enable irq in map_requst() for request allocation, and this spin_lock_irq was a work-around for the case. Now, there is no such case in the device-mapper code, so spin_lock should be enough here. But I'm still using spin_lock_irq for safeness, since there might be some more cases to enable irq during request submission to underlying devices.
I'll remove the _irq in the future after lots of testings.

REVISED PATCH:
This patch fixes the problem that system may stall if target's ->map_rq returns DM_MAPIO_REQUEUE in map_request().
E.g. stall happens on 1 CPU box when a dm-mpath device with queue_if_no_path
bounces between all-paths-down and paths-up on I/O load.

When target's ->map_rq returns DM_MAPIO_REQUEUE, map_request() requeues the request and returns to dm_request_fn(). Then, dm_request_fn() doesn't exit the I/O dispatching loop and continues processing the requeued request again.
This map and requeue loop can be done with interrupt disabled, so 1 CPU system can be stalled if this situation happens.

For example, commands below can stall my 1 CPU box within 1 minute or so:

  # dmsetup table mp
  mp: 0 2097152 multipath 1 queue_if_no_path 0 1 1 service-time 0 1 2 8:144 1 1
  # while true; do dd if=/dev/mapper/mp of=/dev/null bs=1M count=100; done &
  # while true; do \
Quoted 5 lines

dmsetup message mp 0 "fail_path 8:144" \
dmsetup suspend --noflush mp \
dmsetup resume mp \
dmsetup message mp 0 "reinstate_path 8:144" \
done

To fix the problem above, this patch changes dm_request_fn() to exit the I/O dispatching loop once if a request is requeued in map_request().

Cc Alasdair G Kergon <[email protected]>
Cc Greg Kroah-Hartman <[email protected]>
---
 drivers/md/dm.c |   21 +++++++++++++++++----
 1 file changed, 17 insertions(+), 4 deletions(-)
drivers/md/dm.c +17 -4
Index: 2.6.32.8/drivers/md/dm.c
===================================================================
@@ -1487,10 +1487,15 @@ static int dm_prep_fn(struct request_que
 	return BLKPREP_OK;
 }
 
-static void map_request(struct dm_target *ti, struct request *rq,
-			struct mapped_device *md)
+/*
+ * Returns:
+ * 0  : the request has been processed (not requeued)
+ * !0 : the request has been requeued
+ */
+static int map_request(struct dm_target *ti, struct request *rq,
+		       struct mapped_device *md)
 {
-	int r;
+	int r, requeued = 0;
 	struct request *clone = rq->special;
 	struct dm_rq_target_io *tio = clone->end_io_data;
 
@@ -1516,6 +1521,7 @@ static void map_request(struct dm_target
 	case DM_MAPIO_REQUEUE:
 		/* The target wants to requeue the I/O */
 		dm_requeue_unmapped_request(clone);
+		requeued = 1;
 		break;
 	default:
 		if (r > 0) {
@@ -1527,6 +1533,8 @@ static void map_request(struct dm_target
 		dm_kill_unmapped_request(clone, r);
 		break;
 	}
+
+	return requeued;
 }
 
 /*
@@ -1568,12 +1576,17 @@ static void dm_request_fn(struct request
 
 		blk_start_request(rq);
 		spin_unlock(q->queue_lock);
-		map_request(ti, rq, md);
+		if (map_request(ti, rq, md))
+			goto requeued;
+
 		spin_lock_irq(q->queue_lock);
 	}
 
 	goto out;
 
+requeued:
+	spin_lock_irq(q->queue_lock);
+
 plug_and_out:
 	if (!elv_queue_empty(q))
 		/* Some requests still remain, retry later */
#99 lore raw

Re: [Stable-review] [93/93] dm mpath: fix stall when requeueing io

Quoted 19 lines · Kiyoshi Ueda
Quoted 9 lines · Stefan Bader

drivers/md/dm.c static void dm_request_fn(struct request · line 1581

Quoted 9 lines
drivers/md/dm.c +2 -1
@@ -1568,12 +1575,16 @@ static void dm_request_fn(struct request
 
 		blk_start_request(rq);
 		spin_unlock(q->queue_lock);
-		map_request(ti, rq, md);
+		if (map_request(ti, rq, md))
+			goto requeued;
 		spin_lock_irq(q->queue_lock);
 	}

In the current device-mapper code, I would like to go with spin_unlock/lock here.
However, there was a case to enable irq in map_requst() for request allocation, and this spin_lock_irq was a work-around for the case. Now, there is no such case in the device-mapper code, so spin_lock should be enough here. But I'm still using spin_lock_irq for safeness, since there might be some more cases to enable irq during request submission to underlying devices.
I'll remove the _irq in the future after lots of testings.

So, have I understood your reasoning?

- This function (dm_request_fn) is always called with local interrupts disabled. E.g. from generic_unplug_device() or blk_run_queue().

- The 'map_request()' function was found to re-enable interrupts in one case, but that case got fixed.

- The code still uses spin_lock_irq to ensure they remain disabled as protection against there being other cases. This should be changed to spin_lock as a clean-up but you are not aware of any current breakage.

Alasdair

#100 lore raw

Re: [Stable-review] [93/93] dm mpath: fix stall when requeueing io

Hi Alasdair, Linus,

Quoted 26 lines · Alasdair G Kergon
Quoted 19 lines · Kiyoshi Ueda
Quoted 9 lines · Stefan Bader

drivers/md/dm.c static void dm_request_fn(struct request · line 1581

Quoted 9 lines
drivers/md/dm.c +2 -1
@@ -1568,12 +1575,16 @@ static void dm_request_fn(struct request
 
 		blk_start_request(rq);
 		spin_unlock(q->queue_lock);
-		map_request(ti, rq, md);
+		if (map_request(ti, rq, md))
+			goto requeued;
 		spin_lock_irq(q->queue_lock);
 	}

In the current device-mapper code, I would like to go with spin_unlock/lock here.
However, there was a case to enable irq in map_requst() for request allocation, and this spin_lock_irq was a work-around for the case. Now, there is no such case in the device-mapper code, so spin_lock should be enough here. But I'm still using spin_lock_irq for safeness, since there might be some more cases to enable irq during request submission to underlying devices.
I'll remove the _irq in the future after lots of testings.

So, have I understood your reasoning?

- This function (dm_request_fn) is always called with local interrupts disabled. E.g. from generic_unplug_device() or blk_run_queue().

Quoted 6 lines

- The 'map_request()' function was found to re-enable interrupts in one case, but that case got fixed.

- The code still uses spin_lock_irq to ensure they remain disabled as protection against there being other cases. This should be changed to spin_lock as a clean-up but you are not aware of any current breakage.

That's correct.
I think the spin_lock_irq can be changed to spin_lock as a clean-up. But I don't want to break things in this late stage of 2.6.33-rc and/or the stable tree.
So I'll send the clean-up patch for 2.6.34 once I make sure it's ok.

Thanks,
Kiyoshi Ueda

maintainer
#101 lore raw

Re: [Stable-review] [93/93] dm mpath: fix stall when requeueing io

Quoted 36 lines · Kiyoshi Ueda

Hi Alasdair, Linus,

Quoted 26 lines · Alasdair G Kergon
Quoted 19 lines · Kiyoshi Ueda
Quoted 9 lines · Stefan Bader

drivers/md/dm.c static void dm_request_fn(struct request · line 1581

Quoted 9 lines
drivers/md/dm.c +2 -1
@@ -1568,12 +1575,16 @@ static void dm_request_fn(struct request
 
 		blk_start_request(rq);
 		spin_unlock(q->queue_lock);
-		map_request(ti, rq, md);
+		if (map_request(ti, rq, md))
+			goto requeued;
 		spin_lock_irq(q->queue_lock);
 	}

In the current device-mapper code, I would like to go with spin_unlock/lock here.
However, there was a case to enable irq in map_requst() for request allocation, and this spin_lock_irq was a work-around for the case. Now, there is no such case in the device-mapper code, so spin_lock should be enough here. But I'm still using spin_lock_irq for safeness, since there might be some more cases to enable irq during request submission to underlying devices.
I'll remove the _irq in the future after lots of testings.

So, have I understood your reasoning?

- This function (dm_request_fn) is always called with local interrupts disabled. E.g. from generic_unplug_device() or blk_run_queue().

Quoted 6 lines

- The 'map_request()' function was found to re-enable interrupts in one case, but that case got fixed.

- The code still uses spin_lock_irq to ensure they remain disabled as protection against there being other cases. This should be changed to spin_lock as a clean-up but you are not aware of any current breakage.

Quoted 8 lines

That's correct.
I think the spin_lock_irq can be changed to spin_lock as a clean-up. But I don't want to break things in this late stage of 2.6.33-rc and/or the stable tree.
So I'll send the clean-up patch for 2.6.34 once I make sure it's ok.

Thanks,
Kiyoshi Ueda

Yes. Change it in 2.6.34-rc.

Please review the code --- do call graph search over the functions to check that none of them enables interrupts (similar to the search I did to find out that dm_put).

A general note: Never ever try to hide race conditions --- i.e. don't ever try to think like "interrupts should be disabled here but maybe I overlooked something and they are enabled, so I disable them for sure". This doesn't work, it only makes the race condition happen less probably. And the bug that happens less probably is harder to find!

If you want to check that there is no place in map_request() that enables interrupts, use this: BUG_ON(!irqs_disabled()); after map_request() and before spin_lock().

The rationale is this: if you use BUG_ON, and interrupts are accidentally enabled, the user will send you a report with a line number where the crash happened. With the line number, it's pretty easy to find out what happened.

However, if you try to hide the bug and disable interrupts "for sure", the race condition will still happen (i.e. the code that should run with interrupts disabled will be called recursively), but it will happen less likely, and the user will be reporting things like "it locks up once a month" or "it corrupted my filesystem but I can't reproduce it" --- and these things are much-much-much harder to track down than a single BUG().

Mikulas

A message that is not in the archive (sent off-list or not yet received): [email protected]

maintainer
#103 lore raw

Re: [Stable-review] [93/93] dm mpath: fix stall when requeueing io

Quoted 6 lines · Alasdair G Kergon
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
(quoted diff) +2 -1
 		spin_unlock(q->queue_lock);
-		map_request(ti, rq, md);
+		if (map_request(ti, rq, md))
+			goto requeued;
 		spin_lock_irq(q->queue_lock);
Quoted 8 lines
Quoted 8 lines

In the current device-mapper code, I would like to go with spin_unlock/lock here.
However, there was a case to enable irq in map_requst() for request allocation, and this spin_lock_irq was a work-around for the case. Now, there is no such case in the device-mapper code, so spin_lock should be enough here. But I'm still using spin_lock_irq for safeness, since there might be some more cases to enable irq during request submission to underlying devices.

Either map_requst() may enable irqs --- then you should enable them with spin_unlock_irq (then, you'd have to review all the callers of dm_request_fn that they are fine with enabling irqs).

Or map_request() must not enable irqs --- and then there is already a bug and there is no point in using "spin_lock_irq" for safeness, because it doesn't fix the bug (the interrupt may come anyway, before spin_lock_irq).

Quoted 5 lines

I don't understand this.

Are you saying that sometimes this fn is called with local interrupts disabled and other times with them still enabled?

Quoted 7 lines

If they are sometimes left enabled, and the unlock() leaves them alone, then the lock_irq disables them, what piece of code then reenables them?

Surely this code can only be working if local interrupts are always disabled prior to entry?

Alasdair

Another problem:
dm_request_fn can be called in an interrupt context, I scanned it for calling process-context functions and found:

It may call rq_completed (either directly, via dm_request_fn->map_request->dm_kill_unmapped_request->dm_complete_request ->dm_done->dm_end_request->dm_put) or indirectly, when the request is completed from host controller interrupt. And dm_put is a process_context function.

I believe it doesn't cause a real crash, because dm_put is called in dm_blk_close, thus there is always at least one reference. When the device is closed with dm_blk_close, there should be no requests on it.

But it is simply a logic error to call a process-context function from an interrupt context. I'd remove those dm_get/dm_put from request-based-dm --- they are not needed anyway, as long as there are requests, the "mapped_device" structure can't disappear.

You can apply this (in 2.6.34-rc1) to catch all the errorneous users of dm_put.

Mikulas

---

Catch errorneous calls of dm_put from an interrupt context.

---
 drivers/md/dm.c |    1 +
 1 file changed, 1 insertion(+)
drivers/md/dm.c +1 -0
Index: linux-2.6.33-rc8-fast/drivers/md/dm.c
===================================================================
@@ -2188,6 +2188,7 @@ void dm_put(struct mapped_device *md)
 	struct dm_table *map;
 
 	BUG_ON(test_bit(DMF_FREEING, &md->flags));
+	might_sleep();
 
 	if (atomic_dec_and_lock(&md->holders, &_minor_lock)) {
 		map = dm_get_live_table(md);
#104 lore raw

rqdm: bad usage of dm_get/dm_put (Was: Re: dm mpath: fix stall when requeueing io)

Hi Mikulas, Alasdair,

Thank you for spotting this.

Quoted 10 lines · Mikulas Patocka

Another problem:
dm_request_fn can be called in an interrupt context, I scanned it for calling process-context functions and found:

It may call rq_completed (either directly, via dm_request_fn->map_request->dm_kill_unmapped_request->dm_complete_request ->dm_done->dm_end_request->dm_put) or indirectly, when the request is completed from host controller interrupt. And dm_put is a process_context function.

Quoted 8 lines

I believe it doesn't cause a real crash, because dm_put is called in dm_blk_close, thus there is always at least one reference. When the device is closed with dm_blk_close, there should be no requests on it.

But it is simply a logic error to call a process-context function from an interrupt context. I'd remove those dm_get/dm_put from request-based-dm --- they are not needed anyway, as long as there are requests, the "mapped_device" structure can't disappear.

Indeed, we shouldn't use the current dm_put() in any interrupt-context. But the "mapped_device" can disappear in request-based dm while there is a request after all bios complete, so I used dm_get()/dm_put() there. I'll consider another way to prevent the problem without dm_get()/dm_put(). E.g. wait for request completion in dm_put() instead.

Quoted 2 lines

You can apply this (in 2.6.34-rc1) to catch all the errorneous users of dm_put.

<snip>

drivers/md/dm.c void dm_put(struct mapped_device *md) · line 2194

Quoted 8 lines
drivers/md/dm.c +1 -0
@@ -2188,6 +2188,7 @@ void dm_put(struct mapped_device *md)
 	struct dm_table *map;
 
 	BUG_ON(test_bit(DMF_FREEING, &md->flags));
+	might_sleep();
 
 	if (atomic_dec_and_lock(&md->holders, &_minor_lock)) {
 		map = dm_get_live_table(md);

The current request-based dm usually calls dm_put() from softirq context and is warned a lot, so don't apply this patch until I fix the problem above with another way.

Thanks,
Kiyoshi Ueda

maintainer
#105 lore raw

Re: rqdm: bad usage of dm_get/dm_put (Was: Re: dm mpath: fix stall when requeueing io)

Quoted 5 lines

Indeed, we shouldn't use the current dm_put() in any interrupt-context. But the "mapped_device" can disappear in request-based dm while there is a request after all bios complete, so I used dm_get()/dm_put() there. I'll consider another way to prevent the problem without dm_get()/dm_put(). E.g. wait for request completion in dm_put() instead.

How can a request-in-progress exists when all the bios complete and the device is closed?

Mikulas

#106 lore raw

Re: rqdm: bad usage of dm_get/dm_put (Was: Re: dm mpath: fix stall when requeueing io)

Hi Mikulas,

Quoted 8 lines · Mikulas Patocka
Quoted 5 lines

Indeed, we shouldn't use the current dm_put() in any interrupt-context. But the "mapped_device" can disappear in request-based dm while there is a request after all bios complete, so I used dm_get()/dm_put() there. I'll consider another way to prevent the problem without dm_get()/dm_put(). E.g. wait for request completion in dm_put() instead.

How can a request-in-progress exists when all the bios complete and the device is closed?

In the current request-based dm, the device opener can remove the mapped_device while the last request is still completing, because bios in the last request complete first and then the device opener can remove the mapped_device before the last request completes:
CPU0 CPU1
======================================================================
<<INTERRUPT>>
blk_end_request_all(clone_rq)
blk_update_request(clone_rq)

     bio_endio(clone_bio) == end_clone_bio
       blk_update_request(orig_rq)
         bio_endio(orig_bio)
                                                <<I/O completed>>
                                                dm_blk_close()
                                                dev_remove()
                                                  dm_put(md)
                                                    <<Free md>>
   blk_finish_request(clone_rq)
     ....
     dm_end_request(clone_rq)
       free_rq_clone(clone_rq)
       blk_end_request_all(orig_rq)
       rq_completed(md)

So we need a mechanism to defer the md deletion until the last request completes.

Thanks,
Kiyoshi Ueda

maintainer
#107 lore raw

How to unload a module? (Was: [dm-devel] rqdm: bad usage of dm_get/dm_put)

Quoted 36 lines · Kiyoshi Ueda

Hi Mikulas,

Quoted 8 lines · Mikulas Patocka
Quoted 5 lines

Indeed, we shouldn't use the current dm_put() in any interrupt-context. But the "mapped_device" can disappear in request-based dm while there is a request after all bios complete, so I used dm_get()/dm_put() there. I'll consider another way to prevent the problem without dm_get()/dm_put(). E.g. wait for request completion in dm_put() instead.

How can a request-in-progress exists when all the bios complete and the device is closed?

In the current request-based dm, the device opener can remove the mapped_device while the last request is still completing, because bios in the last request complete first and then the device opener can remove the mapped_device before the last request completes:
CPU0 CPU1
======================================================================
<<INTERRUPT>>
blk_end_request_all(clone_rq)
blk_update_request(clone_rq)

     bio_endio(clone_bio) == end_clone_bio
       blk_update_request(orig_rq)
         bio_endio(orig_bio)
                                                <<I/O completed>>
                                                dm_blk_close()
                                                dev_remove()
                                                  dm_put(md)
                                                    <<Free md>>
   blk_finish_request(clone_rq)
     ....
     dm_end_request(clone_rq)
       free_rq_clone(clone_rq)
       blk_end_request_all(orig_rq)
       rq_completed(md)
Quoted 5 lines

So we need a mechanism to defer the md deletion until the last request completes.

Thanks,
Kiyoshi Ueda

Good point ... but I think this problem may happen even in normal non-request based dm.

I don't know what to do with it.

If one thread does:
- bio_endio
and another thread does:
- close the device
- remove the device
- unload module
- then the first thread, after bio_endio, executes non-existing instructions from unloaded module.

Any ideas, how is it solved or how it should be solved?

Module unloading does stop_machine, but AFAIK it waits for all CPUs to exit non-preemtable sections, it doesn't wait for the code to get out of disk request routine...

Mikulas

#108 lore raw

Re: [Stable-review] [93/93] dm mpath: fix stall when requeueing io

In reply to Stefan Bader
Quoted 112 lines · Stefan Bader
Quoted 51 lines · Greg KH

2.6.32-stable review patch. If anyone has any objections, please let us know.

------------------

From: Kiyoshi Ueda <[email protected]>

commit 9eef87da2a8ea4920e0d913ff977cac064b68ee0 upstream.

This patch fixes the problem that system may stall if target's ->map_rq returns DM_MAPIO_REQUEUE in map_request().
E.g. stall happens on 1 CPU box when a dm-mpath device with queue_if_no_path
bounces between all-paths-down and paths-up on I/O load.

When target's ->map_rq returns DM_MAPIO_REQUEUE, map_request() requeues the request and returns to dm_request_fn(). Then, dm_request_fn() doesn't exit the I/O dispatching loop and continues processing the requeued request again.
This map and requeue loop can be done with interrupt disabled, so 1 CPU system can be stalled if this situation happens.

For example, commands below can stall my 1 CPU box within 1 minute or so:

  # dmsetup table mp
  mp: 0 2097152 multipath 1 queue_if_no_path 0 1 1 service-time 0 1 2 8:144 1 1
  # while true; do dd if=/dev/mapper/mp of=/dev/null bs=1M count=100; done &
  # while true; do \
Quoted 5 lines

dmsetup message mp 0 "fail_path 8:144" \
dmsetup suspend --noflush mp \
dmsetup resume mp \
dmsetup message mp 0 "reinstate_path 8:144" \
done

To fix the problem above, this patch changes dm_request_fn() to exit the I/O dispatching loop once if a request is requeued in map_request().

---
 drivers/md/dm.c |   21 ++++++++++++++++-----
 1 file changed, 16 insertions(+), 5 deletions(-)
drivers/md/dm.c +0 -2
@@ -1487,11 +1487,15 @@ static int dm_prep_fn(struct request_que
 	return BLKPREP_OK;
 }
 
-static void map_request(struct dm_target *ti, struct request *rq,
-			struct mapped_device *md)

drivers/md/dm.c static int dm_prep_fn(struct request_que · line 1494

Quoted 10 lines
drivers/md/dm.c +7 -2
+/*
+ * Returns:
+ * 0  : the request has been processed (not requeued)
+ * !0 : the request has been requeued
+ */
+static int map_request(struct dm_target *ti, struct request *clone,
+		       struct mapped_device *md)
 {
-	int r;
-	struct request *clone = rq->special;

This change requires the argument to this function to be a rq->special pointer. This is changed in the map_request function by the following patch:

commit b4324feeae304ae39e631a254d238a7d63be004b

Author: Kiyoshi Ueda <[email protected]>
Date:   Thu Dec 10 23:52:16 2009 +0000

dm: use md pending for in flight IO counting

Quoted 22 lines
drivers/md/dm.c +4 -0
+	int r, requeued = 0;
 	struct dm_rq_target_io *tio = clone->end_io_data;
 
 	/*
@@ -1516,6 +1520,7 @@ static void map_request(struct dm_target
 	case DM_MAPIO_REQUEUE:
 		/* The target wants to requeue the I/O */
 		dm_requeue_unmapped_request(clone);
+		requeued = 1;
 		break;
 	default:
 		if (r > 0) {
@@ -1527,6 +1532,8 @@ static void map_request(struct dm_target
 		dm_kill_unmapped_request(clone, r);
 		break;
 	}
+
+	return requeued;
 }
 
 /*

drivers/md/dm.c static void dm_request_fn(struct request · line 1581

Quoted 8 lines
drivers/md/dm.c +2 -1
@@ -1568,12 +1575,16 @@ static void dm_request_fn(struct request
 
 		blk_start_request(rq);
 		spin_unlock(q->queue_lock);
-		map_request(ti, rq, md);
+		if (map_request(ti, rq, md))
+			goto requeued;
 		spin_lock_irq(q->queue_lock);
 	}

That is the current state of dm_request_function:

                clone = rq->special;
                atomic_inc(&md->pending[rq_data_dir(clone)]);

                spin_unlock(q->queue_lock);
                if (map_request(ti, clone, md))
Quoted 3 lines

While looking over the code I also noticed that the spinlock is dropped with spin_unlock and then reacquired with spin_lock_irq. Isn't the irq version too much in that case? Or was the intention to have interrupts enabled when unlocking?

Ick, thanks for the review. I'll drop this patch for now and fix it up in the next release if needed.

thanks,

greg k-h

#109 lore raw

Re: [Stable-review] [93/93] dm mpath: fix stall when requeueing io

Quoted 18 lines · Greg KH
Quoted 5 lines · Stefan Bader
Quoted 5 lines
Quoted 11 lines

drivers/md/dm.c static void dm_request_fn(struct request · line 1581

Quoted 7 lines
drivers/md/dm.c +2 -1
 		blk_start_request(rq);
 		spin_unlock(q->queue_lock);
-		map_request(ti, rq, md);
+		if (map_request(ti, rq, md))
+			goto requeued;
 		spin_lock_irq(q->queue_lock);
 	}

While looking over the code I also noticed that the spinlock is dropped with spin_unlock and then reacquired with spin_lock_irq. Isn't the irq version too much in that case? Or was the intention to have interrupts enabled when unlocking?

Quoted 2 lines

Ick, thanks for the review. I'll drop this patch for now and fix it up in the next release if needed.

This part of the commentary seems to be relevant for mainline too.

According to git logs, this spin_unlock/spin_lock_irq imbalance goes back to commit cec47e3d4a (last June). Maybe there's some reason for it, but it does look odd.

Ueda-san, Nomura-san, Alasdair? Comments?

Linus