Skip to contentSkip to conversation

-stable review

PATCH In review Chris Wright 7 participants 23 messages Jun 27, 2005 – Jul 1, 2005

Subsystems (8)
lore
Share & tools
lore
#0 lore raw

cover -stable review

This is the start of the stable review cycle for the 2.6.12.2 release. There are 7 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.

These patches are sent out with a number of different people on the Cc: line. If you wish to be a reviewer, please email [email protected] to add your name to the list. If you want to be off the reviewer list, also email us.

Responses should be made by Wed, Jun 29, 23:00 UTC. Anything received after that time, might be too late.

thanks,

the -stable release team (i.e. the ones wearing the joker hat in the corner...)

#1 lore raw

PATCH 1/7 Fix typo in drivers/pci/pci-driver.c

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

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

The git commit 794f5bfa77955c4455f6d72d8b0e2bee25f1ff0c accidentally suffers from a previous typo in that file (',' instead of ';' in end of line). Patch included.

drivers/pci/pci-driver.c +1 -1
Index: linux-2.6/drivers/pci/pci-driver.c
===================================================================
@@ -396,7 +396,7 @@
 	/* FIXME, once all of the existing PCI drivers have been fixed to set
 	 * the pci shutdown function, this test can go away. */
 	if (!drv->driver.shutdown)
-		drv->driver.shutdown = pci_device_shutdown,
+		drv->driver.shutdown = pci_device_shutdown;
 	drv->driver.owner = drv->owner;
 	drv->driver.kobj.ktype = &pci_driver_kobj_type;
 	pci_init_dynids(&drv->dynids);
#2 lore raw

PATCH 2/7 qla2xxx: Pull-down scsi-host-addition to follow board initialization.

In reply to Chris Wright

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

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

Return to previous held-logic of calling scsi_add_host() only after the board has been completely initialized. Also return pci_*() error-codes during probe failure paths.

This also corrects an issue where only lun 0 is being scanned for a given port.

---
commit a1541d5af66d02426655b1498f814c52347dd7d3
tree 02d041e54ebaec744d30ebf6012e305b9673bec0
parent 27198d855abbfc82df69e81b6c8d2f333580114c
author Andrew Vasquez <[email protected]> Thu, 09 Jun 2005 17:21:28 -0700 committer James Bottomley <jejb@mulgrave.(none)> Sat, 11 Jun 2005 13:06:22 -0500

 drivers/scsi/qla2xxx/qla_os.c |   55 +++++++++++++++++++++--------------------
 1 files changed, 28 insertions(+), 27 deletions(-)
drivers/scsi/qla2xxx/qla_os.c +28 -27
@@ -1150,7 +1150,7 @@ iospace_error_exit:
  */
 int qla2x00_probe_one(struct pci_dev *pdev, struct qla_board_info *brd_info)
 {
-	int	ret;
+	int	ret = -ENODEV;
 	device_reg_t __iomem *reg;
 	struct Scsi_Host *host;
 	scsi_qla_host_t *ha;
@@ -1161,7 +1161,7 @@ int qla2x00_probe_one(struct pci_dev *pd
 	fc_port_t *fcport;
 
 	if (pci_enable_device(pdev))
-		return -1;
+		goto probe_out;
 
 	host = scsi_host_alloc(&qla2x00_driver_template,
 	    sizeof(scsi_qla_host_t));
@@ -1183,9 +1183,8 @@ int qla2x00_probe_one(struct pci_dev *pd
 
 	/* Configure PCI I/O space */
 	ret = qla2x00_iospace_config(ha);
-	if (ret != 0) {
-		goto probe_alloc_failed;
-	}
+	if (ret)
+		goto probe_failed;
 
 	/* Sanitize the information from PCI BIOS. */
 	host->irq = pdev->irq;
@@ -1258,23 +1257,10 @@ int qla2x00_probe_one(struct pci_dev *pd
 		qla_printk(KERN_WARNING, ha,
 		    "[ERROR] Failed to allocate memory for adapter\n");
 
-		goto probe_alloc_failed;
+		ret = -ENOMEM;
+		goto probe_failed;
 	}
 
-	pci_set_drvdata(pdev, ha);
-	host->this_id = 255;
-	host->cmd_per_lun = 3;
-	host->unique_id = ha->instance;
-	host->max_cmd_len = MAX_CMDSZ;
-	host->max_channel = ha->ports - 1;
-	host->max_id = ha->max_targets;
-	host->max_lun = ha->max_luns;
-	host->transportt = qla2xxx_transport_template;
-	if (scsi_add_host(host, &pdev->dev))
-		goto probe_alloc_failed;
-
-	qla2x00_alloc_sysfs_attr(ha);
-
 	if (qla2x00_initialize_adapter(ha) &&
 	    !(ha->device_flags & DFLG_NO_CABLE)) {
 
@@ -1285,11 +1271,10 @@ int qla2x00_probe_one(struct pci_dev *pd
 		    "Adapter flags %x.\n",
 		    ha->host_no, ha->device_flags));
 
+		ret = -ENODEV;
 		goto probe_failed;
 	}
 
-	qla2x00_init_host_attr(ha);
-
 	/*
 	 * Startup the kernel thread for this host adapter
 	 */
@@ -1299,17 +1284,26 @@ int qla2x00_probe_one(struct pci_dev *pd
 		qla_printk(KERN_WARNING, ha,
 		    "Unable to start DPC thread!\n");
 
+		ret = -ENODEV;
 		goto probe_failed;
 	}
 	wait_for_completion(&ha->dpc_inited);
 
+	host->this_id = 255;
+	host->cmd_per_lun = 3;
+	host->unique_id = ha->instance;
+	host->max_cmd_len = MAX_CMDSZ;
+	host->max_channel = ha->ports - 1;
+	host->max_lun = MAX_LUNS;
+	host->transportt = qla2xxx_transport_template;
+
 	if (IS_QLA2100(ha) || IS_QLA2200(ha))
 		ret = request_irq(host->irq, qla2100_intr_handler,
 		    SA_INTERRUPT|SA_SHIRQ, ha->brd_info->drv_name, ha);
 	else
 		ret = request_irq(host->irq, qla2300_intr_handler,
 		    SA_INTERRUPT|SA_SHIRQ, ha->brd_info->drv_name, ha);
-	if (ret != 0) {
+	if (ret) {
 		qla_printk(KERN_WARNING, ha,
 		    "Failed to reserve interrupt %d already in use.\n",
 		    host->irq);
@@ -1363,9 +1357,18 @@ int qla2x00_probe_one(struct pci_dev *pd
 		msleep(10);
 	}
 
+	pci_set_drvdata(pdev, ha);
 	ha->flags.init_done = 1;
 	num_hosts++;
 
+	ret = scsi_add_host(host, &pdev->dev);
+	if (ret)
+		goto probe_failed;
+
+	qla2x00_alloc_sysfs_attr(ha);
+
+	qla2x00_init_host_attr(ha);
+
 	qla_printk(KERN_INFO, ha, "\n"
 	    " QLogic Fibre Channel HBA Driver: %s\n"
 	    "  QLogic %s - %s\n"
@@ -1384,9 +1387,6 @@ int qla2x00_probe_one(struct pci_dev *pd
 probe_failed:
 	fc_remove_host(ha->host);
 
-	scsi_remove_host(host);
-
-probe_alloc_failed:
 	qla2x00_free_device(ha);
 
 	scsi_host_put(host);
@@ -1394,7 +1394,8 @@ probe_alloc_failed:
 probe_disable_device:
 	pci_disable_device(pdev);
 
-	return -1;
+probe_out:
+	return ret;
 }
 EXPORT_SYMBOL_GPL(qla2x00_probe_one);
 
#3 lore raw

Re: [02/07] [SCSI] qla2xxx: Pull-down scsi-host-addition to follow board initialization.

Hi Chris, all,

Quoted 2 lines

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

I have. This one patch is rather big and parts of it don't seem to belong to -stable. Can't it be simplified? More below.

Quoted 2 lines

Return to previous held-logic of calling scsi_add_host() only after the board has been completely initialized.

What real bug is it supposed to fix? (I guess some, but this leading comment should give the datails.)

Quoted 1 line

Also return pci_*() error-codes during probe failure paths.

How does this belong to stable please? I don't see this fixing any critical bug.

Quoted 2 lines

This also corrects an issue where only lun 0 is being scanned for a given port.

This OTOH is probably OK.

drivers/scsi/qla2xxx/qla_os.c int qla2x00_probe_one(struct pci_dev *pd · line 1187

Quoted 5 lines
drivers/scsi/qla2xxx/qla_os.c +2 -3
-	if (ret != 0) {
-		goto probe_alloc_failed;
-	}
+	if (ret)
+		goto probe_failed;

This change can be made smaller.

drivers/scsi/qla2xxx/qla_os.c int qla2x00_probe_one(struct pci_dev *pd · line 1306

Quoted 2 lines
drivers/scsi/qla2xxx/qla_os.c +1 -1
-	if (ret != 0) {
+	if (ret) {

This aint -stable material.

Thanks,

Signature · 1 line
Jean Delvare
#4 lore raw

Re: [02/07] [SCSI] qla2xxx: Pull-down scsi-host-addition to follow board initialization.

Quoted 8 lines · Jean Delvare

Hi Chris, all,

Quoted 2 lines

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

I have. This one patch is rather big and parts of it don't seem to belong to -stable. Can't it be simplified? More below.

The threshold for "what belongs in -stable" is a) set too high and b) over-zealously enforced.

Quoted 5 lines
Quoted 2 lines

Return to previous held-logic of calling scsi_add_host() only after the board has been completely initialized.

What real bug is it supposed to fix? (I guess some, but this leading comment should give the datails.)

If that's what was in the patch which went into 2.6.13 then we should be OK with a full backport. If the person who originally raised that patch put unrelated things into a single patch then that's where the problem started.

Bear in mind that there is also risk in only part-applying a patch.

Quoted 4 lines
Quoted 1 line

Also return pci_*() error-codes during probe failure paths.

How does this belong to stable please? I don't see this fixing any critical bug.

But it's obviously safe.

Quoted 4 lines

drivers/scsi/qla2xxx/qla_os.c int qla2x00_probe_one(struct pci_dev *pd · line 1306

Quoted 2 lines
drivers/scsi/qla2xxx/qla_os.c +1 -1
-	if (ret != 0) {
+	if (ret) {

This aint -stable material.

But it's obviously safe. Let's use our brains on these patches and not become beholden to doctrine, OK?

#5 lore raw

Re: [02/07] [SCSI] qla2xxx: Pull-down scsi-host-addition to follow board initialization.

Quoted 2 lines · * Andrew Morton ([email protected]) wrote:

The threshold for "what belongs in -stable" is a) set too high and b) over-zealously enforced.

Do you have things you'd like to see in -stable that didn't make the cut?

Quoted 9 lines
Quoted 5 lines
Quoted 2 lines

Return to previous held-logic of calling scsi_add_host() only after the board has been completely initialized.

What real bug is it supposed to fix? (I guess some, but this leading comment should give the datails.)

If that's what was in the patch which went into 2.6.13 then we should be OK with a full backport. If the person who originally raised that patch put unrelated things into a single patch then that's where the problem started.

Agreed.

Quoted 1 line

Bear in mind that there is also risk in only part-applying a patch.

Yup, if it's only part of the patch, it needs to be re-tested to be sure something important wasn't dropped in the chop up.

Quoted 7 lines
Quoted 4 lines
Quoted 1 line

Also return pci_*() error-codes during probe failure paths.

How does this belong to stable please? I don't see this fixing any critical bug.

But it's obviously safe.

Quoted 7 lines
Quoted 4 lines

drivers/scsi/qla2xxx/qla_os.c int qla2x00_probe_one(struct pci_dev *pd · line 1306

Quoted 2 lines
drivers/scsi/qla2xxx/qla_os.c +1 -1
-	if (ret != 0) {
+	if (ret) {

This aint -stable material.

But it's obviously safe. Let's use our brains on these patches and not become beholden to doctrine, OK?

I agree. The real fix only is 100% preferred, but not at the risk of a patch that's less stable. We've certainly asked for that as the rule of thumb, but it is just that...a rule of thumb.

#6 lore raw

Re: [02/07] [SCSI] qla2xxx: Pull-down scsi-host-addition to follow board initialization.

Quoted 7 lines · Chris Wright
Quoted 2 lines · * Andrew Morton ([email protected]) wrote:

The threshold for "what belongs in -stable" is a) set too high and b) over-zealously enforced.

Do you have things you'd like to see in -stable that didn't make the cut?

Nope - I'm just making vague unsubstantiated accusations ;)

<general handwaving> Given the number of bugs which are present in each release (as evidenced by the amount of stuff we're fixing), there's a hell of a lot of material which _could_ go into -stable.

I suspect some things are slipping through. It's a big job though.

I didn't help much in 2.6.11.x and am paying more attention this time, mainly by being more vigilant looking at the commits list.

2.6.11.x was the first time and things are still getting underway.

#7 lore raw

Re: [02/07] [SCSI] qla2xxx: Pull-down scsi-host-addition to follow board initialization.

Quoted 1 line · * Andrew Morton ([email protected]) wrote:

Nope - I'm just making vague unsubstantiated accusations ;)

Heh ;-)

Quoted 8 lines

<general handwaving> Given the number of bugs which are present in each release (as evidenced by the amount of stuff we're fixing), there's a hell of a lot of material which _could_ go into -stable.

I suspect some things are slipping through. It's a big job though.

I didn't help much in 2.6.11.x and am paying more attention this time, mainly by being more vigilant looking at the commits list.

It would really help if patch author (or committer) thought to send it on, distributing the load...

#8 lore raw

Re: [stable] Re: [02/07] [SCSI] qla2xxx: Pull-down scsi-host-addition to follow board initialization.

In reply to Andrew Morton
Quoted 4 lines · Andrew Morton
Quoted 3 lines · Jean Delvare

Hi Chris, all,

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

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

I have. This one patch is rather big and parts of it don't seem to belong to -stable. Can't it be simplified? More below.

The threshold for "what belongs in -stable" is a) set too high and b) over-zealously enforced.

Hm, are there patches that have been submitted to stable@ that have been rejected for "over-zealous" enforcement? I can't think of any ones recently.

Quoted 11 lines
Quoted 5 lines
Quoted 2 lines

Return to previous held-logic of calling scsi_add_host() only after the board has been completely initialized.

What real bug is it supposed to fix? (I guess some, but this leading comment should give the datails.)

If that's what was in the patch which went into 2.6.13 then we should be OK with a full backport. If the person who originally raised that patch put unrelated things into a single patch then that's where the problem started.

Bear in mind that there is also risk in only part-applying a patch.

I agree. That's why I don't have a problem with this patch, it's better to stay inline with upstream (meaning 2.6.12-git) than diverging. Makes my life easier when I try to figure out if stuff needs to be merged to Linus :)

thanks,

greg k-h

#9 lore raw

Re: [02/07] [SCSI] qla2xxx: Pull-down scsi-host-addition to follow board initialization.

In reply to Andrew Morton

Hi Andrew,

Quoted 2 lines

If the person who originally raised that patch put unrelated things into a single patch then that's where the problem started.

Agreed.

Quoted 1 line

Bear in mind that there is also risk in only part-applying a patch.

If applying only a part of a given patch doesn't sound safe, then I would question the supposed obvious correctness of this patch in the first place.

Some times ago, Alan stated he liked -stable because "its small enough that most of the add on patches people use aren't breaking against it" [1]. I found this a sound statement, but if we now accept non-minimum changes, this won't be true any longer (or at least this will tend to become less true).

Quoted 4 lines
Quoted 1 line

This aint -stable material.

But it's obviously safe. Let's use our brains on these patches and not become beholden to doctrine, OK?

Why did we write down and discuss rules for -stable in the first place then [2]? Of the 9 rules Greg first listed as conditions for a patch to be accepted in -stable, this patch breaks 4 (it is bigger than 100 lines, if fixes more than one thing, including one that doesn't bother people as far as I can see, and it has trivial fixes in it.) So I don't think I am actually splitting hair as you seemed to suggest. I know some of these rules were slightly reworded afterwards, but still.

I reviewed the latest stable series of patches with these rules in mind, trying to help. If the rules have since changed - and it seems they did, then instead of helping, I have been wasting your time, and mine. Where were the new rules discussed? We better have a web page summarizing the current rules for -stable if we want submitters and reviewers to do the right thing.

Thanks.

[1] http://kerneltraffic.org/kernel-traffic/kt20050612_315.html#5
[2] http://kerneltraffic.org/kernel-traffic/kt20050403_303.html#9

Signature · 1 line
Jean Delvare
#10 lore raw

Re: [02/07] [SCSI] qla2xxx: Pull-down scsi-host-addition to follow board initialization.

Quoted 5 lines · Jean Delvare

this patch breaks 4 (it is bigger than 100
lines, if fixes more than one thing, including one that doesn't bother people as far as I can see, and it has trivial fixes in it.) So I don't think I am actually splitting hair as you seemed to suggest. I know some of these rules were slightly reworded afterwards, but still.

Not according to my diffstat, it doesn't:

qla_os.c | 55 ++++++++++++++++++++++++++++---------------------------

 1 files changed, 28 insertions(+), 27 deletions(-)

The basic rationale for this one is that it corrects a late spotted error in a driver update that went into 2.6.12. Without this, the qla2xxx driver crashes for numerous people on loading (linux-scsi has all the bug reports). Since it's clearly a bug fix correcting a major feature failure, that qualifies it for -stable. The two code cleanup corrections look trivial and not worth delaying this over (also, as has been pointed out, they make it identical with the patch applied upstream of 2.6.12, so there wouldn't be any merge issues).

James

#11 lore raw

Re: [02/07] [SCSI] qla2xxx: Pull-down scsi-host-addition to follow board initialization.

Hi James,

Quoted 7 lines
Quoted 1 line

this patch breaks 4 (it is bigger than 100 lines, (...)

Not according to my diffstat, it doesn't:

qla_os.c | 55
++++++++++++++++++++++++++++---------------------------

 1 files changed, 28 insertions(+), 27 deletions(-)

The original rule was including the context, diffstat doesn't. This patch was 136 lines long including the context.

Anyway, it doesn't seem to matter anymore.

Signature · 1 line
Jean Delvare
#12 lore raw

PATCH 3/7 fix remap_pte_range BUG

In reply to Chris Wright

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

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

Out-of-tree user of remap_pfn_range hit kernel BUG at mm/memory.c:1112! It passes an unrounded size to remap_pfn_range, which was okay before 2.6.12, but misses remap_pte_range's new end condition. An audit of all the other ptwalks confirms that this is the only one so exposed.

mm/memory.c +1 -1
@@ -1164,7 +1164,7 @@ int remap_pfn_range(struct vm_area_struc
 {
 	pgd_t *pgd;
 	unsigned long next;
-	unsigned long end = addr + size;
+	unsigned long end = addr + PAGE_ALIGN(size);
 	struct mm_struct *mm = vma->vm_mm;
 	int err;
 
#13 lore raw

PATCH 4/7 e1000: fix spinlock bug

In reply to Chris Wright

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

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

This patch fixes an obvious and nasty bug where we could exit the transmit routine while holding tx_lock.

---

drivers/net/e1000/e1000_main.c +1 -0
diff -urpN -X dontdiff linux-2.6.12-clean/drivers/net/e1000/e1000_main.c linux-2.6.12/drivers/net/e1000/e1000_main.c
@@ -2307,6 +2307,7 @@ e1000_xmit_frame(struct sk_buff *skb, st
 	tso = e1000_tso(adapter, skb);
 	if (tso < 0) {
 		dev_kfree_skb_any(skb);
+		spin_unlock_irqrestore(&adapter->tx_lock, flags);
 		return NETDEV_TX_OK;
 	}
 
#14 lore raw

PATCH 5/7 Add "memory" clobbers to the x86 inline asm of strncmp and friends

In reply to Chris Wright

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

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

From: Linus Torvalds <[email protected]>

Add "memory" clobbers to the x86 inline asm of strncmp and friends

They don't actually clobber memory, but gcc doesn't even know they _read_ memory, so can apparently re-order memory accesses around them.

Which obviously does the wrong thing if the memory access happens to change the memory that the compare function is accessing..

Verified to fix a strange boot problem by Jens Axboe.

---

 include/asm-i386/string.h |   32 ++++++++++++++++++++++----------
 1 files changed, 22 insertions(+), 10 deletions(-)
include/asm-i386/string.h +22 -10
@@ -116,7 +116,8 @@ __asm__ __volatile__(
 	"orb $1,%%al\n"
 	"3:"
 	:"=a" (__res), "=&S" (d0), "=&D" (d1)
-		     :"1" (cs),"2" (ct));
+	:"1" (cs),"2" (ct)
+	:"memory");
 return __res;
 }
 
@@ -138,8 +139,9 @@ __asm__ __volatile__(
 	"3:\tsbbl %%eax,%%eax\n\t"
 	"orb $1,%%al\n"
 	"4:"
-		     :"=a" (__res), "=&S" (d0), "=&D" (d1), "=&c" (d2)
-		     :"1" (cs),"2" (ct),"3" (count));
+	:"=a" (__res), "=&S" (d0), "=&D" (d1), "=&c" (d2)
+	:"1" (cs),"2" (ct),"3" (count)
+	:"memory");
 return __res;
 }
 
@@ -158,7 +160,9 @@ __asm__ __volatile__(
 	"movl $1,%1\n"
 	"2:\tmovl %1,%0\n\t"
 	"decl %0"
-	:"=a" (__res), "=&S" (d0) : "1" (s),"0" (c));
+	:"=a" (__res), "=&S" (d0)
+	:"1" (s),"0" (c)
+	:"memory");
 return __res;
 }
 
@@ -175,7 +179,9 @@ __asm__ __volatile__(
 	"leal -1(%%esi),%0\n"
 	"2:\ttestb %%al,%%al\n\t"
 	"jne 1b"
-	:"=g" (__res), "=&S" (d0), "=&a" (d1) :"0" (0),"1" (s),"2" (c));
+	:"=g" (__res), "=&S" (d0), "=&a" (d1)
+	:"0" (0),"1" (s),"2" (c)
+	:"memory");
 return __res;
 }
 
@@ -189,7 +195,9 @@ __asm__ __volatile__(
 	"scasb\n\t"
 	"notl %0\n\t"
 	"decl %0"
-	:"=c" (__res), "=&D" (d0) :"1" (s),"a" (0), "0" (0xffffffffu));
+	:"=c" (__res), "=&D" (d0)
+	:"1" (s),"a" (0), "0" (0xffffffffu)
+	:"memory");
 return __res;
 }
 
@@ -333,7 +341,9 @@ __asm__ __volatile__(
 	"je 1f\n\t"
 	"movl $1,%0\n"
 	"1:\tdecl %0"
-	:"=D" (__res), "=&c" (d0) : "a" (c),"0" (cs),"1" (count));
+	:"=D" (__res), "=&c" (d0)
+	:"a" (c),"0" (cs),"1" (count)
+	:"memory");
 return __res;
 }
 
@@ -369,7 +379,7 @@ __asm__ __volatile__(
 	"je 2f\n\t"
 	"stosb\n"
 	"2:"
-	: "=&c" (d0), "=&D" (d1)
+	:"=&c" (d0), "=&D" (d1)
 	:"a" (c), "q" (count), "0" (count/4), "1" ((long) s)
 	:"memory");
 return (s);	
@@ -392,7 +402,8 @@ __asm__ __volatile__(
 	"jne 1b\n"
 	"3:\tsubl %2,%0"
 	:"=a" (__res), "=&d" (d0)
-	:"c" (s),"1" (count));
+	:"c" (s),"1" (count)
+	:"memory");
 return __res;
 }
 /* end of additional stuff */
@@ -473,7 +484,8 @@ static inline void * memscan(void * addr
 		"dec %%edi\n"
 		"1:"
 		: "=D" (addr), "=c" (size)
-		: "0" (addr), "1" (size), "a" (c));
+		: "0" (addr), "1" (size), "a" (c)
+		: "memory");
 	return addr;
 }
 
#15 lore raw

Re: [05/07] Add "memory" clobbers to the x86 inline asm of strncmp and friends

Hi Chris,

include/asm-i386/string.h __asm__ __volatile__( · line 146

Quoted 11 lines
include/asm-i386/string.h +3 -2
@@ -138,8 +139,9 @@ __asm__ __volatile__(
 	"3:\tsbbl %%eax,%%eax\n\t"
 	"orb $1,%%al\n"
 	"4:"
-		     :"=a" (__res), "=&S" (d0), "=&D" (d1), "=&c" (d2)
-		     :"1" (cs),"2" (ct),"3" (count));
+	:"=a" (__res), "=&S" (d0), "=&D" (d1), "=&c" (d2)
+	:"1" (cs),"2" (ct),"3" (count)
+	:"memory");
 return __res;
 }

Could be made shorter. As far as I remember, indentation fixes are not -stable material.

include/asm-i386/string.h __asm__ __volatile__( · line 385

Quoted 9 lines
include/asm-i386/string.h +1 -1
@@ -369,7 +379,7 @@ __asm__ __volatile__(
 	"je 2f\n\t"
 	"stosb\n"
 	"2:"
-	: "=&c" (d0), "=&D" (d1)
+	:"=&c" (d0), "=&D" (d1)
 	:"a" (c), "q" (count), "0" (count/4), "1" ((long) s)
 	:"memory");
 return (s);	

Doesn't belong there, unless -stable rules have changed.

Thanks,

Signature · 1 line
Jean Delvare
#16 lore raw

PATCH 6/7 ACPI: Make sure we call acpi_register_gsi() even for default PCI interrupt assignment

In reply to Chris Wright

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

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

From: Linus Torvalds <[email protected]>

ACPI: Make sure we call acpi_register_gsi() even for default PCI interrupt assignment

That's the part that keeps track of the ELCR register, and we want to make sure that the PCI interrupts are properly marked level/low.

---

 drivers/acpi/pci_irq.c |    1 +
 1 files changed, 1 insertion(+)
drivers/acpi/pci_irq.c +1 -0
@@ -435,6 +435,7 @@ acpi_pci_irq_enable (
 		/* Interrupt Line values above 0xF are forbidden */
 		if (dev->irq >= 0 && (dev->irq <= 0xF)) {
 			printk(" - using IRQ %d\n", dev->irq);
+			acpi_register_gsi(dev->irq, ACPI_LEVEL_SENSITIVE, ACPI_ACTIVE_LOW);
 			return_VALUE(0);
 		}
 		else {
#17 lore raw

PATCH 7/7 : Fix two socket hashing bugs.

In reply to Chris Wright

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

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

1) netlink_release() should only decrement the hash entry count if the socket was actually hashed.

This was causing hash->entries to underflow, which resulting in all kinds of troubles.

On 64-bit systems, this would cause the following conditional to erroneously trigger:

	err = -ENOMEM;
	if (BITS_PER_LONG > 32 && unlikely(hash->entries >= UINT_MAX))
		goto err;

2) netlink_autobind() needs to propagate the error return from netlink_insert(). Otherwise, callers will not see the error as they should and thus try to operate on a socket with a zero pid, which is very bad.

However, it should not propagate -EBUSY. If two threads race to autobind the socket, that is fine. This is consistent with the autobind behavior in other protocols.

So bug #1 above, combined with this one, resulted in hangs on netlink_sendmsg() calls to the rtnetlink socket. We'd try to do the user sendmsg() with the socket's pid set to zero, later we do a socket lookup using that pid (via the value we stashed away in NETLINK_CB(skb).pid), but that won't give us the user socket, it will give us the rtnetlink socket. So when we try to wake up the receive queue, we dive back into rtnetlink_rcv() which tries to recursively take the rtnetlink semaphore.

Thanks to Jakub Jelink for providing backtraces. Also, thanks to Herbert Xu for supplying debugging patches to help track this down, and also finding a mistake in an earlier version of this fix.

---

net/netlink/af_netlink.c renamed from net/netlink/af_netlink.c.~1~ +8 -3
@@ -315,8 +315,8 @@
 static void netlink_remove(struct sock *sk)
 {
 	netlink_table_grab();
-	nl_table[sk->sk_protocol].hash.entries--;
-	sk_del_node_init(sk);
+	if (sk_del_node_init(sk))
+		nl_table[sk->sk_protocol].hash.entries--;
 	if (nlk_sk(sk)->groups)
 		__sk_del_bind_node(sk);
 	netlink_table_ungrab();
@@ -429,7 +429,12 @@
 	err = netlink_insert(sk, pid);
 	if (err == -EADDRINUSE)
 		goto retry;
-	return 0;
+
+	/* If 2 threads race to autobind, that is fine.  */
+	if (err == -EBUSY)
+		err = 0;
+
+	return err;
 }
 
 static inline int netlink_capable(struct socket *sock, unsigned int flag) 
#18 lore raw
In reply to Chris Wright
Quoted 2 lines · Chris Wright

Responses should be made by Wed, Jun 29, 23:00 UTC. Anything received after that time, might be too late.

Will the fix for the iptables physdev match go into -stable?

I'd regard the aesthetic objections as secondary in a stable kernel series. If physdev match stays in the kernel, the fix should IMHO go into -stable.

Regards,
Jim

#19 lore raw
Quoted 6 lines · * Jim MacBaine ([email protected]) wrote:
Quoted 2 lines · Chris Wright

Responses should be made by Wed, Jun 29, 23:00 UTC. Anything received after that time, might be too late.

Will the fix for the iptables physdev match go into -stable?

I assume you're referring to this fix:

http://marc.theaimsgroup.com/?l=linux-kernel&m=111936734211687&w=2

If so, I expect it will. Needs to hit mainline first and get pushed over to -stable.