Skip to contentSkip to conversation

fs: remove accidental overflow during wraparound check

PATCH New Sasha Levin 1 participant 8 messages Jul 28, 2024

Subsystems (6)
lore
Share & tools
lore

Sasha Levin

author maintainer
#0 lore raw

PATCH 1/8 fs: remove accidental overflow during wraparound check

From: Justin Stitt <[email protected]>

[ Upstream commit 23cc6ef6fd453b13502caae23130844e7d6ed0fe ]

Running syzkaller with the newly enabled signed integer overflow sanitizer produces this report:

[  195.401651] ------------[ cut here ]------------
[  195.404808] UBSAN: signed-integer-overflow in ../fs/open.c:321:15
[  195.408739] 9223372036854775807 + 562984447377399 cannot be represented in type 'loff_t' (aka 'long long')
[  195.414683] CPU: 1 PID: 703 Comm: syz-executor.0 Not tainted 6.8.0-rc2-00039-g14de58dbe653-dirty #11
[  195.420138] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
[  195.425804] Call Trace:
[  195.427360]  <TASK>
[  195.428791]  dump_stack_lvl+0x93/0xd0
[  195.431150]  handle_overflow+0x171/0x1b0
[  195.433640]  vfs_fallocate+0x459/0x4f0
...
[  195.490053] ------------[ cut here ]------------
[  195.493146] UBSAN: signed-integer-overflow in ../fs/open.c:321:61
[  195.497030] 9223372036854775807 + 562984447377399 cannot be represented in type 'loff_t' (aka 'long long)
[  195.502940] CPU: 1 PID: 703 Comm: syz-executor.0 Not tainted 6.8.0-rc2-00039-g14de58dbe653-dirty #11
[  195.508395] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
[  195.514075] Call Trace:
[  195.515636]  <TASK>
[  195.517000]  dump_stack_lvl+0x93/0xd0
[  195.519255]  handle_overflow+0x171/0x1b0
[  195.521677]  vfs_fallocate+0x4cb/0x4f0
[  195.524033]  __x64_sys_fallocate+0xb2/0xf0

Historically, the signed integer overflow sanitizer did not work in the kernel due to its interaction with -fwrapv but this has since been changed [1] in the newest version of Clang. It was re-enabled in the kernel with Commit 557f8c582a9ba8ab ("ubsan: Reintroduce signed overflow sanitizer").

Let's use the check_add_overflow helper to first verify the addition stays within the bounds of its type (long long); then we can use that sum for the following check.

---
 fs/open.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)
fs/open.c +6 -2
index 0d63c94e1c5e6..dd68725cd7247 100644
@@ -244,6 +244,7 @@ int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
 {
 	struct inode *inode = file_inode(file);
 	long ret;
+	loff_t sum;
 
 	if (offset < 0 || len <= 0)
 		return -EINVAL;
@@ -312,8 +313,11 @@ int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
 	if (!S_ISREG(inode->i_mode) && !S_ISBLK(inode->i_mode))
 		return -ENODEV;
 
-	/* Check for wrap through zero too */
-	if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
+	/* Check for wraparound */
+	if (check_add_overflow(offset, len, &sum))
+		return -EFBIG;
+
+	if (sum > inode->i_sb->s_maxbytes)
 		return -EFBIG;
 
 	if (!file->f_op->fallocate)
Signature · 1 line
2.43.0
author maintainer
#1 lore raw

PATCH 2/8 rcutorture: Fix rcu_torture_fwd_cb_cr() data race

From: "Paul E. McKenney" <[email protected]>

[ Upstream commit 6040072f4774a575fa67b912efe7722874be337b ]

On powerpc systems, spinlock acquisition does not order prior stores against later loads. This means that this statement:

	rfcp->rfc_next = NULL;

Can be reordered to follow this statement:

	WRITE_ONCE(*rfcpp, rfcp);

Which is then a data race with rcu_torture_fwd_prog_cr(), specifically, this statement:

	rfcpn = READ_ONCE(rfcp->rfc_next)

KCSAN located this data race, which represents a real failure on powerpc.

---
 kernel/rcu/rcutorture.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
kernel/rcu/rcutorture.c +1 -1
index 8c45df910763a..c14517912cfaa 100644
@@ -2547,7 +2547,7 @@ static void rcu_torture_fwd_cb_cr(struct rcu_head *rhp)
 	spin_lock_irqsave(&rfp->rcu_fwd_lock, flags);
 	rfcpp = rfp->rcu_fwd_cb_tail;
 	rfp->rcu_fwd_cb_tail = &rfcp->rfc_next;
-	WRITE_ONCE(*rfcpp, rfcp);
+	smp_store_release(rfcpp, rfcp);
 	WRITE_ONCE(rfp->n_launders_cb, rfp->n_launders_cb + 1);
 	i = ((jiffies - rfp->rcu_fwd_startat) / (HZ / FWD_CBS_HIST_DIV));
 	if (i >= ARRAY_SIZE(rfp->n_launders_hist))
Signature · 1 line
2.43.0
author maintainer
#2 lore raw

PATCH 3/8 md: do not delete safemode_timer in mddev_suspend

In reply to Sasha Levin

From: Li Nan <[email protected]>

[ Upstream commit a8768a134518e406d41799a3594aeb74e0889cf7 ]

The deletion of safemode_timer in mddev_suspend() is redundant and potentially harmful now. If timer is about to be woken up but gets deleted, 'in_sync' will remain 0 until the next write, causing array to stay in the 'active' state instead of transitioning to 'clean'.

Commit 0d9f4f135eb6 ("MD: Add del_timer_sync to mddev_suspend (fix nasty panic))" introduced this deletion for dm, because if timer fired after dm is destroyed, the resource which the timer depends on might have been freed.

However, commit 0dd84b319352 ("md: call __md_stop_writes in md_stop") added __md_stop_writes() to md_stop(), which is called before freeing resource. Timer is deleted in __md_stop_writes(), and the origin issue is resolved. Therefore, delete safemode_timer can be removed safely now.

---
 drivers/md/md.c | 1 -
 1 file changed, 1 deletion(-)
drivers/md/md.c +0 -1
index 506c998c0ca59..4183ba70fc143 100644
@@ -489,7 +489,6 @@ void mddev_suspend(struct mddev *mddev)
 	clear_bit_unlock(MD_ALLOW_SB_UPDATE, &mddev->flags);
 	wait_event(mddev->sb_wait, !test_bit(MD_UPDATING_SB, &mddev->flags));
 
-	del_timer_sync(&mddev->safemode_timer);
 	/* restrict memory reclaim I/O during raid array is suspend */
 	mddev->noio_flag = memalloc_noio_save();
 }
Signature · 1 line
2.43.0
author maintainer
#3 lore raw

PATCH 4/8 md/raid5: avoid BUG_ON() while continue reshape after reassembling

In reply to Sasha Levin

From: Yu Kuai <[email protected]>

[ Upstream commit 305a5170dc5cf3d395bb4c4e9239bca6d0b54b49 ]

Currently, mdadm support --revert-reshape to abort the reshape while reassembling, as the test 07revert-grow. However, following BUG_ON() can be triggerred by the test:

kernel BUG at drivers/md/raid5.c:6278!
invalid opcode: 0000 [#1] PREEMPT SMP PTI

irq event stamp: 158985
CPU: 6 PID: 891 Comm: md0_reshape Not tainted 6.9.0-03335-g7592a0b0049a #94
RIP: 0010:reshape_request+0x3f1/0xe60
Call Trace:
 <TASK>
 raid5_sync_request+0x43d/0x550
 md_do_sync+0xb7a/0x2110
 md_thread+0x294/0x2b0
 kthread+0x147/0x1c0
 ret_from_fork+0x59/0x70
 ret_from_fork_asm+0x1a/0x30
 </TASK>

Root cause is that --revert-reshape update the raid_disks from 5 to 4, while reshape position is still set, and after reassembling the array, reshape position will be read from super block, then during reshape the checking of 'writepos' that is caculated by old reshape position will fail.

Fix this panic the easy way first, by converting the BUG_ON() to WARN_ON(), and stop the reshape if checkings fail.

Noted that mdadm must fix --revert-shape as well, and probably md/raid should enhance metadata validation as well, however this means reassemble will fail and there must be user tools to fix the wrong metadata.

---
 drivers/md/raid5.c | 20 +++++++++++++-------
 1 file changed, 13 insertions(+), 7 deletions(-)
drivers/md/raid5.c +13 -7
index ed99b449d8fd4..4315dabd32023 100644
@@ -6316,7 +6316,9 @@ static sector_t reshape_request(struct mddev *mddev, sector_t sector_nr, int *sk
 	safepos = conf->reshape_safe;
 	sector_div(safepos, data_disks);
 	if (mddev->reshape_backwards) {
-		BUG_ON(writepos < reshape_sectors);
+		if (WARN_ON(writepos < reshape_sectors))
+			return MaxSector;
+
 		writepos -= reshape_sectors;
 		readpos += reshape_sectors;
 		safepos += reshape_sectors;
@@ -6334,14 +6336,18 @@ static sector_t reshape_request(struct mddev *mddev, sector_t sector_nr, int *sk
 	 * to set 'stripe_addr' which is where we will write to.
 	 */
 	if (mddev->reshape_backwards) {
-		BUG_ON(conf->reshape_progress == 0);
+		if (WARN_ON(conf->reshape_progress == 0))
+			return MaxSector;
+
 		stripe_addr = writepos;
-		BUG_ON((mddev->dev_sectors &
-			~((sector_t)reshape_sectors - 1))
-		       - reshape_sectors - stripe_addr
-		       != sector_nr);
+		if (WARN_ON((mddev->dev_sectors &
+		    ~((sector_t)reshape_sectors - 1)) -
+		    reshape_sectors - stripe_addr != sector_nr))
+			return MaxSector;
 	} else {
-		BUG_ON(writepos != sector_nr + reshape_sectors);
+		if (WARN_ON(writepos != sector_nr + reshape_sectors))
+			return MaxSector;
+
 		stripe_addr = sector_nr;
 	}
 
Signature · 1 line
2.43.0
author maintainer
#4 lore raw

PATCH 5/8 debugobjects: Annotate racy debug variables

In reply to Sasha Levin

From: Breno Leitao <[email protected]>

[ Upstream commit 5b5baba6222255d29626f63c41f101379ec5400b ]

KCSAN has identified a potential data race in debugobjects, where the global variable debug_objects_maxchain is accessed for both reading and writing simultaneously in separate and parallel data paths. This results in the following splat printed by KCSAN:

  BUG: KCSAN: data-race in debug_check_no_obj_freed / debug_object_activate

write to 0xffffffff847ccfc8 of 4 bytes by task 734 on cpu 41:

  debug_object_activate (lib/debugobjects.c:199 lib/debugobjects.c:564 lib/debugobjects.c:710)
  call_rcu (kernel/rcu/rcu.h:227 kernel/rcu/tree.c:2719 kernel/rcu/tree.c:2838)
  security_inode_free (security/security.c:1626)
  __destroy_inode (./include/linux/fsnotify.h:222 fs/inode.c:287)
  ...
  read to 0xffffffff847ccfc8 of 4 bytes by task 384 on cpu 31:
  debug_check_no_obj_freed (lib/debugobjects.c:1000 lib/debugobjects.c:1019)
  kfree (mm/slub.c:2081 mm/slub.c:4280 mm/slub.c:4390)
  percpu_ref_exit (lib/percpu-refcount.c:147)
  css_free_rwork_fn (kernel/cgroup/cgroup.c:5357)
  ...
  value changed: 0x00000070 -> 0x00000071

The data race is actually harmless as this is just used for debugfs statistics, as all other debug variables.

Annotate all debug variables as racy explicitly, since these variables are known to be racy and harmless.

---
 lib/debugobjects.c | 21 +++++++++++----------
 1 file changed, 11 insertions(+), 10 deletions(-)
lib/debugobjects.c +11 -10
index 5dfa582dbadd2..de7af29937e51 100644
@@ -78,16 +78,17 @@ static bool			obj_freeing;
 /* The number of objs on the global free list */
 static int			obj_nr_tofree;
 
-static int			debug_objects_maxchain __read_mostly;
-static int __maybe_unused	debug_objects_maxchecked __read_mostly;
-static int			debug_objects_fixups __read_mostly;
-static int			debug_objects_warnings __read_mostly;
-static int			debug_objects_enabled __read_mostly
-				= CONFIG_DEBUG_OBJECTS_ENABLE_DEFAULT;
-static int			debug_objects_pool_size __read_mostly
-				= ODEBUG_POOL_SIZE;
-static int			debug_objects_pool_min_level __read_mostly
-				= ODEBUG_POOL_MIN_LEVEL;
+static int __data_racy			debug_objects_maxchain __read_mostly;
+static int __data_racy __maybe_unused	debug_objects_maxchecked __read_mostly;
+static int __data_racy			debug_objects_fixups __read_mostly;
+static int __data_racy			debug_objects_warnings __read_mostly;
+static int __data_racy			debug_objects_enabled __read_mostly
+					= CONFIG_DEBUG_OBJECTS_ENABLE_DEFAULT;
+static int __data_racy			debug_objects_pool_size __read_mostly
+					= ODEBUG_POOL_SIZE;
+static int __data_racy			debug_objects_pool_min_level __read_mostly
+					= ODEBUG_POOL_MIN_LEVEL;
+
 static const struct debug_obj_descr *descr_test  __read_mostly;
 static struct kmem_cache	*obj_cache __read_mostly;
 
Signature · 1 line
2.43.0
author maintainer
#5 lore raw

PATCH 6/8 block: change rq_integrity_vec to respect the iterator

In reply to Sasha Levin

From: Mikulas Patocka <[email protected]>

[ Upstream commit cf546dd289e0f6d2594c25e2fb4e19ee67c6d988 ]

If we allocate a bio that is larger than NVMe maximum request size, attach integrity metadata to it and send it to the NVMe subsystem, the integrity metadata will be corrupted.

Splitting the bio works correctly. The function bio_split will clone the bio, trim the iterator of the first bio and advance the iterator of the second bio.

However, the function rq_integrity_vec has a bug - it returns the first vector of the bio's metadata and completely disregards the metadata iterator that was advanced when the bio was split. Thus, the second bio uses the same metadata as the first bio and this leads to metadata corruption.

This commit changes rq_integrity_vec, so that it calls mp_bvec_iter_bvec instead of returning the first vector. mp_bvec_iter_bvec reads the iterator and uses it to build a bvec for the current position in the iterator.

The "queue_max_integrity_segments(rq->q) > 1" check was removed, because the updated rq_integrity_vec function works correctly with multiple segments.

---
 drivers/nvme/host/pci.c       |  6 +++---
 include/linux/blk-integrity.h | 14 +++++++-------
 2 files changed, 10 insertions(+), 10 deletions(-)
drivers/nvme/host/pci.c +3 -3
index 32e89ea853a47..21712063cbd5a 100644
@@ -873,9 +873,9 @@ static blk_status_t nvme_map_metadata(struct nvme_dev *dev, struct request *req,
 		struct nvme_command *cmnd)
 {
 	struct nvme_iod *iod = blk_mq_rq_to_pdu(req);
+	struct bio_vec bv = rq_integrity_vec(req);
 
-	iod->meta_dma = dma_map_bvec(dev->dev, rq_integrity_vec(req),
-			rq_dma_dir(req), 0);
+	iod->meta_dma = dma_map_bvec(dev->dev, &bv, rq_dma_dir(req), 0);
 	if (dma_mapping_error(dev->dev, iod->meta_dma))
 		return BLK_STS_IOERR;
 	cmnd->rw.metadata = cpu_to_le64(iod->meta_dma);
@@ -1015,7 +1015,7 @@ static __always_inline void nvme_pci_unmap_rq(struct request *req)
 	        struct nvme_iod *iod = blk_mq_rq_to_pdu(req);
 
 		dma_unmap_page(dev->dev, iod->meta_dma,
-			       rq_integrity_vec(req)->bv_len, rq_dma_dir(req));
+			       rq_integrity_vec(req).bv_len, rq_dma_dir(req));
 	}
 
 	if (blk_rq_nr_phys_segments(req))
include/linux/blk-integrity.h +7 -7
index 378b2459efe2d..69f73d0546118 100644
@@ -105,14 +105,13 @@ static inline bool blk_integrity_rq(struct request *rq)
 }
 
 /*
- * Return the first bvec that contains integrity data.  Only drivers that are
- * limited to a single integrity segment should use this helper.
+ * Return the current bvec that contains the integrity data. bip_iter may be
+ * advanced to iterate over the integrity data.
  */
-static inline struct bio_vec *rq_integrity_vec(struct request *rq)
+static inline struct bio_vec rq_integrity_vec(struct request *rq)
 {
-	if (WARN_ON_ONCE(queue_max_integrity_segments(rq->q) > 1))
-		return NULL;
-	return rq->bio->bi_integrity->bip_vec;
+	return mp_bvec_iter_bvec(rq->bio->bi_integrity->bip_vec,
+				 rq->bio->bi_integrity->bip_iter);
 }
 #else /* CONFIG_BLK_DEV_INTEGRITY */
 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
@@ -178,7 +177,8 @@ static inline int blk_integrity_rq(struct request *rq)
 
 static inline struct bio_vec *rq_integrity_vec(struct request *rq)
 {
-	return NULL;
+	/* the optimizer will remove all calls to this function */
+	return (struct bio_vec){ };
 }
 #endif /* CONFIG_BLK_DEV_INTEGRITY */
 #endif /* _LINUX_BLK_INTEGRITY_H */
Signature · 1 line
2.43.0
author maintainer
#6 lore raw

PATCH 7/8 rcu: Fix rcu_barrier() VS post CPUHP_TEARDOWN_CPU invocation

In reply to Sasha Levin

From: Frederic Weisbecker <[email protected]>

[ Upstream commit 55d4669ef1b76823083caecfab12a8bd2ccdcf64 ]

When rcu_barrier() calls rcu_rdp_cpu_online() and observes a CPU off rnp->qsmaskinitnext, it means that all accesses from the offline CPU preceding the CPUHP_TEARDOWN_CPU are visible to RCU barrier, including callbacks expiration and counter updates.

However interrupts can still fire after stop_machine() re-enables interrupts and before rcutree_report_cpu_dead(). The related accesses happening between CPUHP_TEARDOWN_CPU and rnp->qsmaskinitnext clearing are _NOT_ guaranteed to be seen by rcu_barrier() without proper ordering, especially when callbacks are invoked there to the end, making rcutree_migrate_callback() bypass barrier_lock.

The following theoretical race example can make rcu_barrier() hang:

CPU 0 CPU 1
----- -----
//cpu_down()
smpboot_park_threads()
//ksoftirqd is parked now

<IRQ>

rcu_sched_clock_irq()
invoke_rcu_core()
do_softirq()
rcu_core()

      rcu_do_batch()
         // callback storm
         // rcu_do_batch() returns
         // before completing all
         // of them
   // do_softirq also returns early because of
   // timeout. It defers to ksoftirqd but
   // it's parked
</IRQ>
stop_machine()
   take_cpu_down()
                                                    rcu_barrier()
                                                        spin_lock(barrier_lock)
                                                        // observes rcu_segcblist_n_cbs(&rdp->cblist) != 0
<IRQ>
do_softirq()
   rcu_core()
      rcu_do_batch()
         //completes all pending callbacks
         //smp_mb() implied _after_ callback number dec
</IRQ>

rcutree_report_cpu_dead()
rnp->qsmaskinitnext &= ~rdp->grpmask;

rcutree_migrate_callback()
// no callback, early return without locking
// barrier_lock

                                                        //observes !rcu_rdp_cpu_online(rdp)
                                                        rcu_barrier_entrain()
                                                           rcu_segcblist_entrain()
                                                              // Observe rcu_segcblist_n_cbs(rsclp) == 0
                                                              // because no barrier between reading
                                                              // rnp->qsmaskinitnext and rsclp->len
                                                              rcu_segcblist_add_len()
                                                                 smp_mb__before_atomic()
                                                                 // will now observe the 0 count and empty
                                                                 // list, but too late, we enqueue regardless
                                                                 WRITE_ONCE(rsclp->len, rsclp->len + v);
                                                        // ignored barrier callback
                                                        // rcu barrier stall...

This could be solved with a read memory barrier, enforcing the message passing between rnp->qsmaskinitnext and rsclp->len, matching the full memory barrier after rsclp->len addition in rcu_segcblist_add_len() performed at the end of rcu_do_batch().

However the rcu_barrier() is complicated enough and probably doesn't need too many more subtleties. CPU down is a slowpath and the barrier_lock seldom contended. Solve the issue with unconditionally locking the barrier_lock on rcutree_migrate_callbacks(). This makes sure that either rcu_barrier() sees the empty queue or its entrained callback will be migrated.

---
 kernel/rcu/tree.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)
kernel/rcu/tree.c +7 -3
index 61f9503a5fe9c..cd6144cea5a1a 100644
@@ -4391,11 +4391,15 @@ void rcutree_migrate_callbacks(int cpu)
 	struct rcu_data *rdp = per_cpu_ptr(&rcu_data, cpu);
 	bool needwake;
 
-	if (rcu_rdp_is_offloaded(rdp) ||
-	    rcu_segcblist_empty(&rdp->cblist))
-		return;  /* No callbacks to migrate. */
+	if (rcu_rdp_is_offloaded(rdp))
+		return;
 
 	raw_spin_lock_irqsave(&rcu_state.barrier_lock, flags);
+	if (rcu_segcblist_empty(&rdp->cblist)) {
+		raw_spin_unlock_irqrestore(&rcu_state.barrier_lock, flags);
+		return;  /* No callbacks to migrate. */
+	}
+
 	WARN_ON_ONCE(rcu_rdp_cpu_online(rdp));
 	rcu_barrier_entrain(rdp);
 	my_rdp = this_cpu_ptr(&rcu_data);
Signature · 1 line
2.43.0
author maintainer
#7 lore raw

PATCH 8/8 clocksource/drivers/sh_cmt: Address race condition for clock events

In reply to Sasha Levin

From: Niklas Söderlund <[email protected]>

[ Upstream commit db19d3aa77612983a02bd223b3f273f896b243cf ]

There is a race condition in the CMT interrupt handler. In the interrupt handler the driver sets a driver private flag, FLAG_IRQCONTEXT. This flag is used to indicate any call to set_next_event() should not be directly propagated to the device, but instead cached. This is done as the interrupt handler itself reprograms the device when needed before it completes and this avoids this operation to take place twice.

It is unclear why this design was chosen, my suspicion is to allow the struct clock_event_device.event_handler callback, which is called while the FLAG_IRQCONTEXT is set, can update the next event without having to write to the device twice.

Unfortunately there is a race between when the FLAG_IRQCONTEXT flag is set and later cleared where the interrupt handler have already started to write the next event to the device. If set_next_event() is called in this window the value is only cached in the driver but not written. This leads to the board to misbehave, or worse lockup and produce a splat.

   rcu: INFO: rcu_preempt detected stalls on CPUs/tasks:
   rcu:     0-...!: (0 ticks this GP) idle=f5e0/0/0x0 softirq=519/519 fqs=0 (false positive?)
   rcu:     (detected by 1, t=6502 jiffies, g=-595, q=77 ncpus=2)
   Sending NMI from CPU 1 to CPUs 0:
   NMI backtrace for cpu 0
   CPU: 0 PID: 0 Comm: swapper/0 Not tainted 6.10.0-rc5-arm64-renesas-00019-g74a6f86eaf1c-dirty #20
   Hardware name: Renesas Salvator-X 2nd version board based on r8a77965 (DT)
   pstate: 60000005 (nZCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
   pc : tick_check_broadcast_expired+0xc/0x40
   lr : cpu_idle_poll.isra.0+0x8c/0x168
   sp : ffff800081c63d70
   x29: ffff800081c63d70 x28: 00000000580000c8 x27: 00000000bfee5610
   x26: 0000000000000027 x25: 0000000000000000 x24: 0000000000000000
   x23: ffff00007fbb9100 x22: ffff8000818f1008 x21: ffff8000800ef07c
   x20: ffff800081c79ec0 x19: ffff800081c70c28 x18: 0000000000000000
   x17: 0000000000000000 x16: 0000000000000000 x15: 0000ffffc2c717d8
   x14: 0000000000000000 x13: ffff000009c18080 x12: ffff8000825f7fc0
   x11: 0000000000000000 x10: ffff8000818f3cd4 x9 : 0000000000000028
   x8 : ffff800081c79ec0 x7 : ffff800081c73000 x6 : 0000000000000000
   x5 : 0000000000000000 x4 : ffff7ffffe286000 x3 : 0000000000000000
   x2 : ffff7ffffe286000 x1 : ffff800082972900 x0 : ffff8000818f1008
   Call trace:
    tick_check_broadcast_expired+0xc/0x40
    do_idle+0x9c/0x280
    cpu_startup_entry+0x34/0x40
    kernel_init+0x0/0x11c
    do_one_initcall+0x0/0x260
    __primary_switched+0x80/0x88
   rcu: rcu_preempt kthread timer wakeup didn't happen for 6501 jiffies! g-595 f0x0 RCU_GP_WAIT_FQS(5) ->state=0x402
   rcu:     Possible timer handling issue on cpu=0 timer-softirq=262
   rcu: rcu_preempt kthread starved for 6502 jiffies! g-595 f0x0 RCU_GP_WAIT_FQS(5) ->state=0x402 ->cpu=0
   rcu:     Unless rcu_preempt kthread gets sufficient CPU time, OOM is now expected behavior.
   rcu: RCU grace-period kthread stack dump:
   task:rcu_preempt     state:I stack:0     pid:15    tgid:15    ppid:2      flags:0x00000008
   Call trace:
    __switch_to+0xbc/0x100
    __schedule+0x358/0xbe0
    schedule+0x48/0x148
    schedule_timeout+0xc4/0x138
    rcu_gp_fqs_loop+0x12c/0x764
    rcu_gp_kthread+0x208/0x298
    kthread+0x10c/0x110
    ret_from_fork+0x10/0x20

The design have been part of the driver since it was first merged in early 2009. It becomes increasingly harder to trigger the issue the older kernel version one tries. It only takes a few boots on v6.10-rc5, while hundreds of boots are needed to trigger it on v5.10.

Close the race condition by using the CMT channel lock for the two competing sections. The channel lock was added to the driver after its initial design.

---
 drivers/clocksource/sh_cmt.c | 13 ++++++++++++-
 1 file changed, 12 insertions(+), 1 deletion(-)
drivers/clocksource/sh_cmt.c +12 -1
index 7b952aa52c0b9..7a2b83157bf5f 100644
@@ -529,6 +529,7 @@ static void sh_cmt_set_next(struct sh_cmt_channel *ch, unsigned long delta)
 static irqreturn_t sh_cmt_interrupt(int irq, void *dev_id)
 {
 	struct sh_cmt_channel *ch = dev_id;
+	unsigned long flags;
 
 	/* clear flags */
 	sh_cmt_write_cmcsr(ch, sh_cmt_read_cmcsr(ch) &
@@ -559,6 +560,8 @@ static irqreturn_t sh_cmt_interrupt(int irq, void *dev_id)
 
 	ch->flags &= ~FLAG_SKIPEVENT;
 
+	raw_spin_lock_irqsave(&ch->lock, flags);
+
 	if (ch->flags & FLAG_REPROGRAM) {
 		ch->flags &= ~FLAG_REPROGRAM;
 		sh_cmt_clock_event_program_verify(ch, 1);
@@ -571,6 +574,8 @@ static irqreturn_t sh_cmt_interrupt(int irq, void *dev_id)
 
 	ch->flags &= ~FLAG_IRQCONTEXT;
 
+	raw_spin_unlock_irqrestore(&ch->lock, flags);
+
 	return IRQ_HANDLED;
 }
 
@@ -781,12 +786,18 @@ static int sh_cmt_clock_event_next(unsigned long delta,
 				   struct clock_event_device *ced)
 {
 	struct sh_cmt_channel *ch = ced_to_sh_cmt(ced);
+	unsigned long flags;
 
 	BUG_ON(!clockevent_state_oneshot(ced));
+
+	raw_spin_lock_irqsave(&ch->lock, flags);
+
 	if (likely(ch->flags & FLAG_IRQCONTEXT))
 		ch->next_match_value = delta - 1;
 	else
-		sh_cmt_set_next(ch, delta - 1);
+		__sh_cmt_set_next(ch, delta - 1);
+
+	raw_spin_unlock_irqrestore(&ch->lock, flags);
 
 	return 0;
 }
Signature · 1 line
2.43.0