Skip to contentSkip to conversation

possible recursive locking detected (blkdev_open)

BUG? Rolf Eike Beer 5 participants 7 messages Aug 9, 2006 – Aug 21, 2006 VFS

lore
Share & tools
lore
#0 lore raw

=============================================
[ INFO: possible recursive locking detected ]
---------------------------------------------
parted/7929 is trying to acquire lock:
(&bdev->bd_mutex){--..}, at: [<c105eb8d>] __blkdev_put+0x1e/0x13c

but task is already holding lock:
(&bdev->bd_mutex){--..}, at: [<c105eec6>] do_open+0x72/0x3a8

other info that might help us debug this:
1 lock held by parted/7929:
#0: (&bdev->bd_mutex){--..}, at: [<c105eec6>] do_open+0x72/0x3a8

stack backtrace:
 [<c1003aad>] show_trace_log_lvl+0x58/0x15b
 [<c100495f>] show_trace+0xd/0x10
 [<c1004979>] dump_stack+0x17/0x1a
 [<c102dee5>] __lock_acquire+0x753/0x99c
 [<c102e3b0>] lock_acquire+0x4a/0x6a
 [<c1204501>] mutex_lock_nested+0xc8/0x20c
 [<c105eb8d>] __blkdev_put+0x1e/0x13c
 [<c105ecc4>] blkdev_put+0xa/0xc
 [<c105f18a>] do_open+0x336/0x3a8
 [<c105f21b>] blkdev_open+0x1f/0x4c
 [<c1057b40>] __dentry_open+0xc7/0x1aa
 [<c1057c91>] nameidata_to_filp+0x1c/0x2e
 [<c1057cd1>] do_filp_open+0x2e/0x35
 [<c1057dd7>] do_sys_open+0x38/0x68
 [<c1057e33>] sys_open+0x16/0x18
 [<c1002845>] sysenter_past_esp+0x56/0x8d
DWARF2 unwinder stuck at sysenter_past_esp+0x56/0x8d
Leftover inexact backtrace:
 [<c100495f>] show_trace+0xd/0x10
 [<c1004979>] dump_stack+0x17/0x1a
 [<c102dee5>] __lock_acquire+0x753/0x99c
 [<c102e3b0>] lock_acquire+0x4a/0x6a
 [<c1204501>] mutex_lock_nested+0xc8/0x20c
 [<c105eb8d>] __blkdev_put+0x1e/0x13c
 [<c105ecc4>] blkdev_put+0xa/0xc
 [<c105f18a>] do_open+0x336/0x3a8
 [<c105f21b>] blkdev_open+0x1f/0x4c
 [<c1057b40>] __dentry_open+0xc7/0x1aa
 [<c1057c91>] nameidata_to_filp+0x1c/0x2e
 [<c1057cd1>] do_filp_open+0x2e/0x35
 [<c1057dd7>] do_sys_open+0x38/0x68
 [<c1057e33>] sys_open+0x16/0x18
 [<c1002845>] sysenter_past_esp+0x56/0x8d
#1 lore raw
Quoted 38 lines · On Wed, 9 Aug 2006 07:57:31 +0200 Rolf Eike Beer <[email protected]> wrote:

=============================================
[ INFO: possible recursive locking detected ]
---------------------------------------------
parted/7929 is trying to acquire lock:
(&bdev->bd_mutex){--..}, at: [<c105eb8d>] __blkdev_put+0x1e/0x13c

but task is already holding lock:
(&bdev->bd_mutex){--..}, at: [<c105eec6>] do_open+0x72/0x3a8

other info that might help us debug this:
1 lock held by parted/7929:
#0: (&bdev->bd_mutex){--..}, at: [<c105eec6>] do_open+0x72/0x3a8

stack backtrace:
 [<c1003aad>] show_trace_log_lvl+0x58/0x15b
 [<c100495f>] show_trace+0xd/0x10
 [<c1004979>] dump_stack+0x17/0x1a
 [<c102dee5>] __lock_acquire+0x753/0x99c
 [<c102e3b0>] lock_acquire+0x4a/0x6a
 [<c1204501>] mutex_lock_nested+0xc8/0x20c
 [<c105eb8d>] __blkdev_put+0x1e/0x13c
 [<c105ecc4>] blkdev_put+0xa/0xc
 [<c105f18a>] do_open+0x336/0x3a8
 [<c105f21b>] blkdev_open+0x1f/0x4c
 [<c1057b40>] __dentry_open+0xc7/0x1aa
 [<c1057c91>] nameidata_to_filp+0x1c/0x2e
 [<c1057cd1>] do_filp_open+0x2e/0x35
 [<c1057dd7>] do_sys_open+0x38/0x68
 [<c1057e33>] sys_open+0x16/0x18
 [<c1002845>] sysenter_past_esp+0x56/0x8d
DWARF2 unwinder stuck at sysenter_past_esp+0x56/0x8d
Leftover inexact backtrace:
 [<c100495f>] show_trace+0xd/0x10
 [<c1004979>] dump_stack+0x17/0x1a
 [<c102dee5>] __lock_acquire+0x753/0x99c
 [<c102e3b0>] lock_acquire+0x4a/0x6a
 [<c1204501>] mutex_lock_nested+0xc8/0x20c
 [<c105eb8d>] __blkdev_put+0x1e/0x13c
 [<c105ecc4>] blkdev_put+0xa/0xc
Quoted 9 lines
 [<c105f18a>] do_open+0x336/0x3a8
 [<c105f21b>] blkdev_open+0x1f/0x4c
 [<c1057b40>] __dentry_open+0xc7/0x1aa
 [<c1057c91>] nameidata_to_filp+0x1c/0x2e
 [<c1057cd1>] do_filp_open+0x2e/0x35
 [<c1057dd7>] do_sys_open+0x38/0x68
 [<c1057e33>] sys_open+0x16/0x18
 [<c1002845>] sysenter_past_esp+0x56/0x8d

kernel version?

#2 lore raw
Quoted 9 lines · Andrew Morton

On Wed, 9 Aug 2006 07:57:31 +0200

Quoted 6 lines · Rolf Eike Beer

=============================================
[ INFO: possible recursive locking detected ]
---------------------------------------------
parted/7929 is trying to acquire lock:
(&bdev->bd_mutex){--..}, at: [<c105eb8d>] __blkdev_put+0x1e/0x13c

Quoted 4 lines
Quoted 2 lines

but task is already holding lock:
(&bdev->bd_mutex){--..}, at: [<c105eec6>] do_open+0x72/0x3a8

kernel version?

compiled from git 2006-08-03 16:02 (+0200), it's -rc3 and a bit.

Eike

#3 lore raw
In reply to Andrew Morton
Quoted 8 lines · Andrew Morton
Quoted 3 lines · On Wed, 9 Aug 2006 07:57:31 +0200 Rolf Eike Beer <[email protected]> wrote:

=============================================
[ INFO: possible recursive locking detected ]
---------------------------------------------

kernel version?

This question comes up time after time when we get lockdep reports. Lets do the same thing we do for oopses - print out the version in the report. It's an extra line of output though. We could tack it on the end of the
INFO: lines, but that screws up Ingo's pretty output.

kernel/lockdep.c renamed from kernel/lockdep.c~ +16 -0
@@ -36,6 +36,7 @@
 #include <linux/stacktrace.h>
 #include <linux/debug_locks.h>
 #include <linux/irqflags.h>
+#include <linux/utsname.h>
 
 #include <asm/sections.h>
@@ -524,6 +524,9 @@ print_circular_bug_header(struct lock_li
 
 	printk("\n=======================================================\n");
 	printk(  "[ INFO: possible circular locking dependency detected ]\n");
+	printk(  "%s %.*s\n", system_utsname.release,
+		(int)strcspn(system_utsname.version, " "),
+		system_utsname.version);
 	printk(  "-------------------------------------------------------\n");
 	printk("%s/%d is trying to acquire lock:\n",
 		curr->comm, curr->pid);
@@ -705,6 +708,9 @@ print_bad_irq_dependency(struct task_str
 	printk("\n======================================================\n");
 	printk(  "[ INFO: %s-safe -> %s-unsafe lock order detected ]\n",
 		irqclass, irqclass);
+	printk(  "%s %.*s\n", system_utsname.release,
+		(int)strcspn(system_utsname.version, " "),
+		system_utsname.version);
 	printk(  "------------------------------------------------------\n");
 	printk("%s/%d [HC%u[%lu]:SC%u[%lu]:HE%u:SE%u] is trying to acquire:\n",
 		curr->comm, curr->pid,
@@ -786,6 +792,9 @@ print_deadlock_bug(struct task_struct *c
 
 	printk("\n=============================================\n");
 	printk(  "[ INFO: possible recursive locking detected ]\n");
+	printk(  "%s %.*s\n", system_utsname.release,
+		(int)strcspn(system_utsname.version, " "),
+		system_utsname.version);
 	printk(  "---------------------------------------------\n");
 	printk("%s/%d is trying to acquire lock:\n",
 		curr->comm, curr->pid);
@@ -1368,6 +1377,9 @@ print_irq_inversion_bug(struct task_stru
 
 	printk("\n=========================================================\n");
 	printk(  "[ INFO: possible irq lock inversion dependency detected ]\n");
+	printk(  "%s %.*s\n", system_utsname.release,
+		(int)strcspn(system_utsname.version, " "),
+		system_utsname.version);
 	printk(  "---------------------------------------------------------\n");
 	printk("%s/%d just changed the state of lock:\n",
 		curr->comm, curr->pid);
@@ -1462,6 +1474,9 @@ print_usage_bug(struct task_struct *curr
 
 	printk("\n=================================\n");
 	printk(  "[ INFO: inconsistent lock state ]\n");
+	printk(  "%s %.*s\n", system_utsname.release,
+		(int)strcspn(system_utsname.version, " "),
+		system_utsname.version);
 	printk(  "---------------------------------\n");
 
 	printk("inconsistent {%s} -> {%s} usage.\n",
Signature · 1 line
http://www.codemonkey.org.uk
#4 lore raw

kernel/lockdep.c print_circular_bug_header(struct lock_li

Quoted 28 lines · On Wed, 9 Aug 2006 13:56:42 -0400 Dave Jones <[email protected]> wrote:
Quoted 8 lines · Andrew Morton
Quoted 3 lines · On Wed, 9 Aug 2006 07:57:31 +0200 Rolf Eike Beer <[email protected]> wrote:

=============================================
[ INFO: possible recursive locking detected ]
---------------------------------------------

kernel version?

This question comes up time after time when we get lockdep reports. Lets do the same thing we do for oopses - print out the version in the report. It's an extra line of output though. We could tack it on the end of the
INFO: lines, but that screws up Ingo's pretty output.

kernel/lockdep.c renamed from kernel/lockdep.c~ +1 -0
@@ -36,6 +36,7 @@
 #include <linux/stacktrace.h>
 #include <linux/debug_locks.h>
 #include <linux/irqflags.h>
+#include <linux/utsname.h>
 
 #include <asm/sections.h>
@@ -524,6 +524,9 @@ print_circular_bug_header(struct lock_li

hm, corrupted patch. Needed a blank line before the @@ line,

kernel/lockdep.c print_circular_bug_header(struct lock_li · line 528

Quoted 5 lines
kernel/lockdep.c renamed from kernel/lockdep.c~ +3 -0
 	printk("\n=======================================================\n");
 	printk(  "[ INFO: possible circular locking dependency detected ]\n");
+	printk(  "%s %.*s\n", system_utsname.release,
+		(int)strcspn(system_utsname.version, " "),
+		system_utsname.version);

argh. Every time someone adds one of these I get to go and fix up namespaces-utsname-*.patch again.

So I did it as below:

kernel/lockdep.c renamed from kernel/lockdep.c~lockdep-print-kernel-version +13 -0
@@ -36,6 +36,7 @@
 #include <linux/stacktrace.h>
 #include <linux/debug_locks.h>
 #include <linux/irqflags.h>
+#include <linux/utsname.h>
 
 #include <asm/sections.h>
 
@@ -508,6 +509,13 @@ print_circular_bug_entry(struct lock_lis
 	return 0;
 }
 
+static void print_kernel_version(void)
+{
+	printk("%s %.*s\n", system_utsname.release,
+		(int)strcspn(system_utsname.version, " "),
+		system_utsname.version);
+}
+
 /*
  * When a circular dependency is detected, print the
  * header first:
@@ -524,6 +532,7 @@ print_circular_bug_header(struct lock_li
 
 	printk("\n=======================================================\n");
 	printk(  "[ INFO: possible circular locking dependency detected ]\n");
+	print_kernel_version();
 	printk(  "-------------------------------------------------------\n");
 	printk("%s/%d is trying to acquire lock:\n",
 		curr->comm, curr->pid);
@@ -705,6 +714,7 @@ print_bad_irq_dependency(struct task_str
 	printk("\n======================================================\n");
 	printk(  "[ INFO: %s-safe -> %s-unsafe lock order detected ]\n",
 		irqclass, irqclass);
+	print_kernel_version();
 	printk(  "------------------------------------------------------\n");
 	printk("%s/%d [HC%u[%lu]:SC%u[%lu]:HE%u:SE%u] is trying to acquire:\n",
 		curr->comm, curr->pid,
@@ -786,6 +796,7 @@ print_deadlock_bug(struct task_struct *c
 
 	printk("\n=============================================\n");
 	printk(  "[ INFO: possible recursive locking detected ]\n");
+	print_kernel_version();
 	printk(  "---------------------------------------------\n");
 	printk("%s/%d is trying to acquire lock:\n",
 		curr->comm, curr->pid);
@@ -1368,6 +1379,7 @@ print_irq_inversion_bug(struct task_stru
 
 	printk("\n=========================================================\n");
 	printk(  "[ INFO: possible irq lock inversion dependency detected ]\n");
+	print_kernel_version();
 	printk(  "---------------------------------------------------------\n");
 	printk("%s/%d just changed the state of lock:\n",
 		curr->comm, curr->pid);
@@ -1462,6 +1474,7 @@ print_usage_bug(struct task_struct *curr
 
 	printk("\n=================================\n");
 	printk(  "[ INFO: inconsistent lock state ]\n");
+	print_kernel_version();
 	printk(  "---------------------------------\n");
 
 	printk("inconsistent {%s} -> {%s} usage.\n",

_

#5 lore raw
In reply to Rolf Eike Beer

(partial CC list from commit 663d440eaa496db903cc58be04b9b602ba45e43b)

Quoted 21 lines · Rolf Eike Beer

=============================================
[ INFO: possible recursive locking detected ]
---------------------------------------------
parted/7929 is trying to acquire lock:
(&bdev->bd_mutex){--..}, at: [<c105eb8d>] __blkdev_put+0x1e/0x13c

but task is already holding lock:
(&bdev->bd_mutex){--..}, at: [<c105eec6>] do_open+0x72/0x3a8

other info that might help us debug this:
1 lock held by parted/7929:
#0: (&bdev->bd_mutex){--..}, at: [<c105eec6>] do_open+0x72/0x3a8

stack backtrace:
 [<c1003aad>] show_trace_log_lvl+0x58/0x15b
 [<c100495f>] show_trace+0xd/0x10
 [<c1004979>] dump_stack+0x17/0x1a
 [<c102dee5>] __lock_acquire+0x753/0x99c
 [<c102e3b0>] lock_acquire+0x4a/0x6a
 [<c1204501>] mutex_lock_nested+0xc8/0x20c
 [<c105eb8d>] __blkdev_put+0x1e/0x13c
 [<c105ecc4>] blkdev_put+0xa/0xc
Quoted 8 lines
 [<c105f18a>] do_open+0x336/0x3a8
 [<c105f21b>] blkdev_open+0x1f/0x4c
 [<c1057b40>] __dentry_open+0xc7/0x1aa
 [<c1057c91>] nameidata_to_filp+0x1c/0x2e
 [<c1057cd1>] do_filp_open+0x2e/0x35
 [<c1057dd7>] do_sys_open+0x38/0x68
 [<c1057e33>] sys_open+0x16/0x18
 [<c1002845>] sysenter_past_esp+0x56/0x8d

OK, I'm having a look here; its all new to me so bear with me.

blkdev_open() calls
do_open(bdev, ...,BD_MUTEX_NORMAL) and takes

    mutex_lock_nested(&bdev->bd_mutex, BD_MUTEX_NORMAL)

then something fails, and we're thrown to:

out_first: where

    if (bdev != bdev->bd_contains)
      blkdev_put(bdev->bd_contains) which is
        __blkdev_put(bdev->bd_contains, BD_MUTEX_NORMAL) which does
          mutex_lock_nested(&bdev->bd_contains->bd_mutex, BD_MUTEX_NORMAL) <--- lockdep trigger

When going to out_first, dbev->bd_contains is either bdev or whole, and since we take the branch it must be whole. So it seems to me the following patch would be the right one:

---
 fs/block_dev.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
fs/block_dev.c +1 -1
Index: linux-2.6/fs/block_dev.c
===================================================================
@@ -980,7 +980,7 @@ out_first:
 	bdev->bd_disk = NULL;
 	bdev->bd_inode->i_data.backing_dev_info = &default_backing_dev_info;
 	if (bdev != bdev->bd_contains)
-		blkdev_put(bdev->bd_contains);
+		__blkdev_put(bdev->bd_contains, BD_MUTEX_WHOLE);
 	bdev->bd_contains = NULL;
 	put_disk(disk);
 	module_put(owner);
Acked-by
#6 lore raw
Quoted 8 lines · On Friday August 18, [email protected] wrote:

blkdev_open() calls
do_open(bdev, ...,BD_MUTEX_NORMAL) and takes

    mutex_lock_nested(&bdev->bd_mutex, BD_MUTEX_NORMAL)

then something fails, and we're thrown to:

out_first: where

Quoted 8 lines
    if (bdev != bdev->bd_contains)
      blkdev_put(bdev->bd_contains) which is
        __blkdev_put(bdev->bd_contains, BD_MUTEX_NORMAL) which does
          mutex_lock_nested(&bdev->bd_contains->bd_mutex, BD_MUTEX_NORMAL) <--- lockdep trigger

When going to out_first, dbev->bd_contains is either bdev or whole, and since we take the branch it must be whole. So it seems to me the following patch would be the right one:

Looks sensible to me.

Quoted 2 lines

NeilBrown

Quoted 27 lines
---
 fs/block_dev.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Index: linux-2.6/fs/block_dev.c
===================================================================

fs/block_dev.c +1 -2
@@ -980,7 +980,7 @@ out_first:
 	bdev->bd_disk = NULL;
 	bdev->bd_inode->i_data.backing_dev_info = &default_backing_dev_info;
 	if (bdev != bdev->bd_contains)
-		blkdev_put(bdev->bd_contains);
+		__blkdev_put(bdev->bd_contains, BD_MUTEX_WHOLE);
 	bdev->bd_contains = NULL;
 	put_disk(disk);
 	module_put(owner);
 
 
 
 
 
-

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/