Recently started seeing a kernel oops when a module tries removing a
memory mapped sysfs bin_attribute. On closer investigation the root
cause seems to be kernfs_release_file() trying to call
kernfs_op.release() callback that's NULL for such sysfs
bin_attributes. The oops occurs when kernfs_release_file() is called from
kernfs_drain_open_files() to cleanup any open handles with active
memory mappings.
The patch fixes this by checking for flag KERNFS_HAS_RELEASE before
calling kernfs_release_file() in function kernfs_drain_open_files().
On ppc64-le arch with cxl module the oops back-trace is of the
form below:
[ 861.381126] Unable to handle kernel paging request for instruction fetch
[ 861.381360] Faulting instruction address: 0x00000000
[ 861.381428] Oops: Kernel access of bad area, sig: 11 [#1]
....
[ 861.382481] NIP: 0000000000000000 LR: c000000000362c60 CTR:
0000000000000000
....
Call Trace:
[c000000f1680b750] [c000000000362c34] kernfs_drain_open_files+0x104/0x1d0 (unreliable)
[c000000f1680b790] [c00000000035fa00] __kernfs_remove+0x260/0x2c0
[c000000f1680b820] [c000000000360da0] kernfs_remove_by_name_ns+0x60/0xe0
[c000000f1680b8b0] [c0000000003638f4] sysfs_remove_bin_file+0x24/0x40
[c000000f1680b8d0] [c00000000062a164] device_remove_bin_file+0x24/0x40
[c000000f1680b8f0] [d000000009b7b22c] cxl_sysfs_afu_remove+0x144/0x170 [cxl]
[c000000f1680b940] [d000000009b7c7e4] cxl_remove+0x6c/0x1a0 [cxl]
[c000000f1680b990] [c00000000052f694] pci_device_remove+0x64/0x110
[c000000f1680b9d0] [c0000000006321d4] device_release_driver_internal+0x1f4/0x2b0
[c000000f1680ba20] [c000000000525cb0] pci_stop_bus_device+0xa0/0xd0
[c000000f1680ba60] [c000000000525e80] pci_stop_and_remove_bus_device+0x20/0x40
[c000000f1680ba90] [c00000000004a6c4] pci_hp_remove_devices+0x84/0xc0
[c000000f1680bad0] [c00000000004a688] pci_hp_remove_devices+0x48/0xc0
[c000000f1680bb10] [c0000000009dfda4] eeh_reset_device+0xb0/0x290
[c000000f1680bbb0] [c000000000032b4c] eeh_handle_normal_event+0x47c/0x530
[c000000f1680bc60] [c000000000032e64] eeh_handle_event+0x174/0x350
[c000000f1680bd10] [c000000000033228] eeh_event_handler+0x1e8/0x1f0
[c000000f1680bdc0] [c0000000000d384c] kthread+0x14c/0x190
[c000000f1680be30] [c00000000000b5a0] ret_from_kernel_thread+0x5c/0xbc
Fixes: f83f3c515654("kernfs: fix locking around kernfs_ops->release()
callback")
Signed-off-by: Vaibhav Jain <[email protected]>
---
fs/kernfs/file.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/fs/kernfs/file.c b/fs/kernfs/file.c
index 8e4dc7a..ac2dfe0 100644
--- a/fs/kernfs/file.c
+++ b/fs/kernfs/file.c
@@ -809,7 +809,8 @@ void kernfs_drain_open_files(struct kernfs_node *kn)
if (kn->flags & KERNFS_HAS_MMAP)
unmap_mapping_range(inode->i_mapping, 0, 0, 1);
- kernfs_release_file(kn, of);
+ if (kn->flags & KERNFS_HAS_RELEASE)
+ kernfs_release_file(kn, of);
}
mutex_unlock(&kernfs_open_file_mutex);
--
2.9.3
On Tue, Mar 14, 2017 at 08:17:00AM +0530, Vaibhav Jain wrote:
> Recently started seeing a kernel oops when a module tries removing a
> memory mapped sysfs bin_attribute. On closer investigation the root
> cause seems to be kernfs_release_file() trying to call
> kernfs_op.release() callback that's NULL for such sysfs
> bin_attributes. The oops occurs when kernfs_release_file() is called from
> kernfs_drain_open_files() to cleanup any open handles with active
> memory mappings.
>
> The patch fixes this by checking for flag KERNFS_HAS_RELEASE before
> calling kernfs_release_file() in function kernfs_drain_open_files().
>
> On ppc64-le arch with cxl module the oops back-trace is of the
> form below:
> [ 861.381126] Unable to handle kernel paging request for instruction fetch
> [ 861.381360] Faulting instruction address: 0x00000000
> [ 861.381428] Oops: Kernel access of bad area, sig: 11 [#1]
> ....
> [ 861.382481] NIP: 0000000000000000 LR: c000000000362c60 CTR:
> 0000000000000000
> ....
> Call Trace:
> [c000000f1680b750] [c000000000362c34] kernfs_drain_open_files+0x104/0x1d0 (unreliable)
> [c000000f1680b790] [c00000000035fa00] __kernfs_remove+0x260/0x2c0
> [c000000f1680b820] [c000000000360da0] kernfs_remove_by_name_ns+0x60/0xe0
> [c000000f1680b8b0] [c0000000003638f4] sysfs_remove_bin_file+0x24/0x40
> [c000000f1680b8d0] [c00000000062a164] device_remove_bin_file+0x24/0x40
> [c000000f1680b8f0] [d000000009b7b22c] cxl_sysfs_afu_remove+0x144/0x170 [cxl]
> [c000000f1680b940] [d000000009b7c7e4] cxl_remove+0x6c/0x1a0 [cxl]
> [c000000f1680b990] [c00000000052f694] pci_device_remove+0x64/0x110
> [c000000f1680b9d0] [c0000000006321d4] device_release_driver_internal+0x1f4/0x2b0
> [c000000f1680ba20] [c000000000525cb0] pci_stop_bus_device+0xa0/0xd0
> [c000000f1680ba60] [c000000000525e80] pci_stop_and_remove_bus_device+0x20/0x40
> [c000000f1680ba90] [c00000000004a6c4] pci_hp_remove_devices+0x84/0xc0
> [c000000f1680bad0] [c00000000004a688] pci_hp_remove_devices+0x48/0xc0
> [c000000f1680bb10] [c0000000009dfda4] eeh_reset_device+0xb0/0x290
> [c000000f1680bbb0] [c000000000032b4c] eeh_handle_normal_event+0x47c/0x530
> [c000000f1680bc60] [c000000000032e64] eeh_handle_event+0x174/0x350
> [c000000f1680bd10] [c000000000033228] eeh_event_handler+0x1e8/0x1f0
> [c000000f1680bdc0] [c0000000000d384c] kthread+0x14c/0x190
> [c000000f1680be30] [c00000000000b5a0] ret_from_kernel_thread+0x5c/0xbc
>
> Fixes: f83f3c515654("kernfs: fix locking around kernfs_ops->release()
> callback")
> Signed-off-by: Vaibhav Jain <[email protected]>
> ---
> fs/kernfs/file.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/fs/kernfs/file.c b/fs/kernfs/file.c
> index 8e4dc7a..ac2dfe0 100644
> --- a/fs/kernfs/file.c
> +++ b/fs/kernfs/file.c
> @@ -809,7 +809,8 @@ void kernfs_drain_open_files(struct kernfs_node *kn)
> if (kn->flags & KERNFS_HAS_MMAP)
> unmap_mapping_range(inode->i_mapping, 0, 0, 1);
>
> - kernfs_release_file(kn, of);
> + if (kn->flags & KERNFS_HAS_RELEASE)
> + kernfs_release_file(kn, of);
> }
>
> mutex_unlock(&kernfs_open_file_mutex);
> --
> 2.9.3
Tejun, want to take this through your tree, or at the least, give me an
ack for this?
thanks,
greg k-h
On Tue, Mar 14, 2017 at 08:17:00AM +0530, Vaibhav Jain wrote:
> Recently started seeing a kernel oops when a module tries removing a
> memory mapped sysfs bin_attribute. On closer investigation the root
> cause seems to be kernfs_release_file() trying to call
> kernfs_op.release() callback that's NULL for such sysfs
> bin_attributes. The oops occurs when kernfs_release_file() is called from
> kernfs_drain_open_files() to cleanup any open handles with active
> memory mappings.
>
> The patch fixes this by checking for flag KERNFS_HAS_RELEASE before
> calling kernfs_release_file() in function kernfs_drain_open_files().
>
> On ppc64-le arch with cxl module the oops back-trace is of the
> form below:
> [ 861.381126] Unable to handle kernel paging request for instruction fetch
> [ 861.381360] Faulting instruction address: 0x00000000
> [ 861.381428] Oops: Kernel access of bad area, sig: 11 [#1]
> ....
> [ 861.382481] NIP: 0000000000000000 LR: c000000000362c60 CTR:
> 0000000000000000
> ....
> Call Trace:
> [c000000f1680b750] [c000000000362c34] kernfs_drain_open_files+0x104/0x1d0 (unreliable)
> [c000000f1680b790] [c00000000035fa00] __kernfs_remove+0x260/0x2c0
> [c000000f1680b820] [c000000000360da0] kernfs_remove_by_name_ns+0x60/0xe0
> [c000000f1680b8b0] [c0000000003638f4] sysfs_remove_bin_file+0x24/0x40
> [c000000f1680b8d0] [c00000000062a164] device_remove_bin_file+0x24/0x40
> [c000000f1680b8f0] [d000000009b7b22c] cxl_sysfs_afu_remove+0x144/0x170 [cxl]
> [c000000f1680b940] [d000000009b7c7e4] cxl_remove+0x6c/0x1a0 [cxl]
> [c000000f1680b990] [c00000000052f694] pci_device_remove+0x64/0x110
> [c000000f1680b9d0] [c0000000006321d4] device_release_driver_internal+0x1f4/0x2b0
> [c000000f1680ba20] [c000000000525cb0] pci_stop_bus_device+0xa0/0xd0
> [c000000f1680ba60] [c000000000525e80] pci_stop_and_remove_bus_device+0x20/0x40
> [c000000f1680ba90] [c00000000004a6c4] pci_hp_remove_devices+0x84/0xc0
> [c000000f1680bad0] [c00000000004a688] pci_hp_remove_devices+0x48/0xc0
> [c000000f1680bb10] [c0000000009dfda4] eeh_reset_device+0xb0/0x290
> [c000000f1680bbb0] [c000000000032b4c] eeh_handle_normal_event+0x47c/0x530
> [c000000f1680bc60] [c000000000032e64] eeh_handle_event+0x174/0x350
> [c000000f1680bd10] [c000000000033228] eeh_event_handler+0x1e8/0x1f0
> [c000000f1680bdc0] [c0000000000d384c] kthread+0x14c/0x190
> [c000000f1680be30] [c00000000000b5a0] ret_from_kernel_thread+0x5c/0xbc
>
> Fixes: f83f3c515654("kernfs: fix locking around kernfs_ops->release()
> callback")
> Signed-off-by: Vaibhav Jain <[email protected]>
Acked-by: Tejun Heo <[email protected]>
Thanks.
--
tejun
Hello, Greg.
On Tue, Mar 14, 2017 at 11:08:29AM +0800, Greg Kroah-Hartman wrote:
> Tejun, want to take this through your tree, or at the least, give me an
> ack for this?
Just acked. I think going through your tree is better for this one.
Thanks!
--
tejun
On Thu, Mar 16, 2017 at 05:14:30PM -0400, Tejun Heo wrote:
> Hello, Greg.
>
> On Tue, Mar 14, 2017 at 11:08:29AM +0800, Greg Kroah-Hartman wrote:
> > Tejun, want to take this through your tree, or at the least, give me an
> > ack for this?
>
> Just acked. I think going through your tree is better for this one.
Ok, will take it in my tree, thanks.
greg k-h