BUG: kernel NULL pointer dereference, address: 0000000000000008 #32
Comments
|
Linux r6 5.13.0-21-generic #21-Ubuntu SMP Tue Oct 19 08:59:28 UTC 2021 x86_64 x86_64 x86_64 GNU/Linux |
|
Could you please validate latest 5.13.x ? there was a similar problem which was fixed/backported in there. |
|
thanks, but not sure where to find a 5.13 newer than 5.13.21? |
|
So you mean 5.13.19 also has this problem?? |
|
yes, the same as 5.13.21. |
|
I seem to have the same issue: $ uname -a |
|
The problem is floating between the kernels, still no root cause found. |
Are you sure, it’s inconsistently present? I get consistently a null deref in aq_ring_rx_fill for the latest 3 stable/longterm kernel releases and mainline¹ when hibernating with the atlantic network interface up. I’ve found a line to be consistently referenced in vmcore dumps of null pointer dereferences. @koo5: I see resume log messages in your log. Was that resume from hibernation and was the device turned off by that hibernation or stayed it on? For me the hibernation immediately resumes without switching the PC off. After that “resume” the null pointer deref occurs. There is no indication from kernel messages, why the system resumes immediately after writing the image. I captured vmcores for the latter three of ¹5.10.109, 5.15.32, 5.17.1, 5.18-rc1 with crashkernel and panic_on_oops and tracked it down to this line in 5.18-rc1:
dmesg-5.18-rc1.log The other logs from Redhat’s crash utility look similar, even with the same offset. I could not retrieve the vmcore dump from 5.10.109 because crashkernel could not reinitialize USB and SATA when booting. Note, kernel is tainted by an amdgpu warning. Probably some module/builtin incompatibility. Needed it builtin, because crashkernel would leave me blind otherwise. In contrary to that, here are dmesg logs of one hibernation with the oldest AQC107 supporting longterm kernels: Bisecting is not easily possible unfortunately, because all the latest, but unsupported minor versions between 5.4 and 5.10 do not compile with my toolchain. Tried also older gcc versions and -Wno-error, but to no avail. I hope you can just build the kernel and reproduce the issue, because I can’t provide the core dumps. They contain my LUKS key in kmem and since it is distributed all over it 1, I’d rather not try to find and scramble it. If you need any more info like other gdb/crash outputs or the other used kernel configs (basically an olddefconfig of the one provided), leave me a note. OS and build environment Portage 3.0.30 (python 3.9.9-final-0, default/linux/amd64/17.1/hardened, gcc-11.2.1, glibc-2.34-r10, 5.17.1-gentoo-r1-x86_64 x86_64) System uname: Linux-5.17.1-gentoo-r1-x86_64-x86_64-AMD_Ryzen_Threadripper_3990X_64-Core_Processor-with-glibc2.34 |
intel-lab-lkp
pushed a commit
to intel-lab-lkp/linux
that referenced
this issue
Apr 18, 2022
From 6c4cd8210835489da84bf4ee5049945dc0f2c986 Mon Sep 17 00:00:00 2001
Date: Mon, 18 Apr 2022 00:20:01 +0200
This will reset deeply on freeze and thaw instead of suspend and
resume and prevent null pointer dereferences of the uninitialized ring
0 buffer while thawing.
The impact is an indefinitely hanging kernel. You can't switch
consoles after this and the only possible user interaction is SysRq.
BUG: kernel NULL pointer dereference
RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic]
aq_vec_init+0x85/0xe0 [atlantic]
aq_nic_init+0xf7/0x1d0 [atlantic]
atl_resume_common+0x4f/0x100 [atlantic]
pci_pm_thaw+0x42/0xa0
resolves in aq_ring.o to
```
0000000000000ae0 <aq_ring_rx_fill>:
{
/* ... */
baf: 48 8b 43 08 mov 0x8(%rbx),%rax
buff->flags = 0U; /* buff is NULL */
```
The bug has been present since the introduction of the new pm code in
8aaa112 ("net: atlantic: refactoring pm logic") and was hidden
until 8ce8427 ("net: atlantic: changes for multi-TC support"),
which refactored the aq_vec_{free,alloc} functions into
aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The
original functions just always reinitialized the buffers on S3/S4. If
the interface is down before freezing, the bug does not occur. It does
not matter, whether the initrd contains and loads the module before
thawing.
So the fix is to invert the boolean parameter deep in all pm function
calls, which was clearly intended to be set like that.
First report was on Github [1], which you have to guess from the
resume logs in the posted dmesg snippet. Recently I posted one on
Bugzilla [2], since I did not have an AQC device so far.
#regzbot introduced: 8ce8427
#regzbot from: koo5 <[email protected]>
#regzbot monitor: Aquantia/AQtion#32
Fixes: 8aaa112 ("net: atlantic: refactoring pm logic")
Link: Aquantia/AQtion#32 [1]
Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2]
Cc: [email protected]
Reported-by: koo5 <[email protected]>
Signed-off-by: Manuel Ullmann <[email protected]>
|
I’ve posted a patch for this, since I found the root cause. The deep parameter for pm functions in aq_pci_func.c was inverted for all calls, so suspend/resume did a deep reset, while freeze/thaw did a normal reset. This was present since the introduction of the new pm API usage in 5.4-rc5 with 8aaa112a57c1 ("net: atlantic: refactoring pm logic"), however it had no effect until aq_vec_{free,alloc} were split in 8ce84271697a("net: atlantic: changes for multi-TC support"), leading to ring buffers not being reinitialized on non-deep pm calls. The latter commit is found by bisect. @koo5: Thank you for reporting this; I mentioned you in the Reported-by tag. |
ammarfaizi2
pushed a commit
to ammarfaizi2/linux-block
that referenced
this issue
Apr 18, 2022
This will reset deeply on freeze and thaw instead of suspend and
resume and prevent null pointer dereferences of the uninitialized ring
0 buffer while thawing.
The impact is an indefinitely hanging kernel. You can't switch
consoles after this and the only possible user interaction is SysRq.
BUG: kernel NULL pointer dereference
RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic]
aq_vec_init+0x85/0xe0 [atlantic]
aq_nic_init+0xf7/0x1d0 [atlantic]
atl_resume_common+0x4f/0x100 [atlantic]
pci_pm_thaw+0x42/0xa0
resolves in aq_ring.o to
```
0000000000000ae0 <aq_ring_rx_fill>:
{
/* ... */
baf: 48 8b 43 08 mov 0x8(%rbx),%rax
buff->flags = 0U; /* buff is NULL */
```
The bug has been present since the introduction of the new pm code in
8aaa112 ("net: atlantic: refactoring pm logic") and was hidden
until 8ce8427 ("net: atlantic: changes for multi-TC support"),
which refactored the aq_vec_{free,alloc} functions into
aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The
original functions just always reinitialized the buffers on S3/S4. If
the interface is down before freezing, the bug does not occur. It does
not matter, whether the initrd contains and loads the module before
thawing.
So the fix is to invert the boolean parameter deep in all pm function
calls, which was clearly intended to be set like that.
First report was on Github [1], which you have to guess from the
resume logs in the posted dmesg snippet. Recently I posted one on
Bugzilla [2], since I did not have an AQC device so far.
#regzbot introduced: 8ce8427
#regzbot from: koo5 <[email protected]>
#regzbot monitor: Aquantia/AQtion#32
Fixes: 8aaa112 ("net: atlantic: refactoring pm logic")
Link: Aquantia/AQtion#32 [1]
Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2]
Cc: [email protected]
Reported-by: koo5 <[email protected]>
Signed-off-by: Manuel Ullmann <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
|
Hm, taking a closer look at koo5’s reported trace, this is a null dereference on resume, so it could not be related, unless the deep reset on suspend/resume causes null derefs for other devices. Getting some retesting from the reporters might help to clear this up. @koo5, @eitch: Could you please retest any of the patches, applied against the kernel version, indicated by filename? Thanks in advance. 5.10.111-invert-deep-par.patch.txt |
ammarfaizi2
pushed a commit
to ammarfaizi2/linux-block
that referenced
this issue
Apr 25, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
ammarfaizi2
pushed a commit
to ammarfaizi2/linux-block
that referenced
this issue
Apr 25, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
ammarfaizi2
pushed a commit
to ammarfaizi2/linux-block
that referenced
this issue
Apr 25, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
ammarfaizi2
pushed a commit
to ammarfaizi2/linux-block
that referenced
this issue
Apr 25, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
ammarfaizi2
pushed a commit
to ammarfaizi2/linux-block
that referenced
this issue
Apr 25, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
ammarfaizi2
pushed a commit
to ammarfaizi2/linux-block
that referenced
this issue
Apr 25, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
ammarfaizi2
pushed a commit
to ammarfaizi2/linux-block
that referenced
this issue
Apr 25, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
ammarfaizi2
pushed a commit
to ammarfaizi2/linux-block
that referenced
this issue
Apr 25, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
ammarfaizi2
pushed a commit
to ammarfaizi2/linux-block
that referenced
this issue
Apr 25, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
ammarfaizi2
pushed a commit
to ammarfaizi2/linux-block
that referenced
this issue
Apr 25, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
ammarfaizi2
pushed a commit
to ammarfaizi2/linux-block
that referenced
this issue
Apr 25, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
ammarfaizi2
pushed a commit
to ammarfaizi2/linux-block
that referenced
this issue
Apr 25, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
ammarfaizi2
pushed a commit
to ammarfaizi2/linux-block
that referenced
this issue
Apr 25, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
ammarfaizi2
pushed a commit
to ammarfaizi2/linux-block
that referenced
this issue
Apr 25, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
ammarfaizi2
pushed a commit
to ammarfaizi2/linux-block
that referenced
this issue
Apr 25, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
ammarfaizi2
pushed a commit
to ammarfaizi2/linux-block
that referenced
this issue
Apr 25, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
ammarfaizi2
pushed a commit
to ammarfaizi2/linux-block
that referenced
this issue
Apr 25, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
ammarfaizi2
pushed a commit
to ammarfaizi2/linux-block
that referenced
this issue
Apr 27, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
ammarfaizi2
pushed a commit
to ammarfaizi2/linux-block
that referenced
this issue
Apr 27, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
ammarfaizi2
pushed a commit
to ammarfaizi2/linux-block
that referenced
this issue
Apr 27, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
ammarfaizi2
pushed a commit
to ammarfaizi2/linux-block
that referenced
this issue
Apr 27, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
ammarfaizi2
pushed a commit
to ammarfaizi2/linux-block
that referenced
this issue
Apr 27, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
ammarfaizi2
pushed a commit
to ammarfaizi2/linux-block
that referenced
this issue
Apr 27, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
Whissi
pushed a commit
to Whissi/linux-stable
that referenced
this issue
Apr 27, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
ammarfaizi2
pushed a commit
to ammarfaizi2/linux-block
that referenced
this issue
Apr 27, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
ammarfaizi2
pushed a commit
to ammarfaizi2/linux-fork
that referenced
this issue
Apr 27, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
ammarfaizi2
pushed a commit
to ammarfaizi2/linux-fork
that referenced
this issue
Apr 27, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
ammarfaizi2
pushed a commit
to ammarfaizi2/linux-block
that referenced
this issue
Apr 27, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
ammarfaizi2
pushed a commit
to ammarfaizi2/linux-block
that referenced
this issue
Apr 27, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
woodsts
pushed a commit
to woodsts/linux-stable
that referenced
this issue
Apr 27, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
woodsts
pushed a commit
to woodsts/linux-stable
that referenced
this issue
Apr 27, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
ammarfaizi2
pushed a commit
to ammarfaizi2/linux-block
that referenced
this issue
Apr 27, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
|
Much appreciated. Sorry i did not get to testing your patches yet. But:
it was from an attempted suspend, the suspend never works, my machine resumes immediately. |
Never mind. This is patched now in stable as of 5.17.5, 5.15.36 and 5.10.113. Could you try any of those kernel versions?
Oh, I see. I’ve also seen that behaviour with hibernation shifting between null deref during hibernation (causing immediate thaw) and after hibernation while bisecting, so Igor’s impression might be valid. I could not reproduce the kernel oops before hibernation with the latest stable patches before my patch though, e.g. 5.10.110. Which kernel version do you use? I never had issues with suspend in 5.15.32 (dist-kernel), any 5.16 version (can’t recall them) and 5.17. Since I was investigating hibernation, I did not try suspend in older kernel releases, but I can give them a try. If you can still reproduce this, knowing the reproducing kernel version would be much appreciated. :) Side note: |
psndna88
pushed a commit
to psndna88/AGNi-xanmod_x86-64
that referenced
this issue
May 3, 2022
commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
|
@labre Hi, the change associated with this appears to have caused a regression. See https://bugzilla.kernel.org/show_bug.cgi?id=215949 |
|
@amarshall Thanks for your bug report and I’m sorry for the regression. I screwed up suspend testing by kexec-rebooting in an unpatched kernel. This is handled in |
RezaT4795
pushed a commit
to MahStudio/WSLackware-kernel
that referenced
this issue
May 31, 2022
commit cbe6c3a8f8f4315b96e46e1a1c70393c06d95a4c upstream.
This will reset deeply on freeze and thaw instead of suspend and
resume and prevent null pointer dereferences of the uninitialized ring
0 buffer while thawing.
The impact is an indefinitely hanging kernel. You can't switch
consoles after this and the only possible user interaction is SysRq.
BUG: kernel NULL pointer dereference
RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic]
aq_vec_init+0x85/0xe0 [atlantic]
aq_nic_init+0xf7/0x1d0 [atlantic]
atl_resume_common+0x4f/0x100 [atlantic]
pci_pm_thaw+0x42/0xa0
resolves in aq_ring.o to
```
0000000000000ae0 <aq_ring_rx_fill>:
{
/* ... */
baf: 48 8b 43 08 mov 0x8(%rbx),%rax
buff->flags = 0U; /* buff is NULL */
```
The bug has been present since the introduction of the new pm code in
8aaa112 ("net: atlantic: refactoring pm logic") and was hidden
until 8ce8427 ("net: atlantic: changes for multi-TC support"),
which refactored the aq_vec_{free,alloc} functions into
aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The
original functions just always reinitialized the buffers on S3/S4. If
the interface is down before freezing, the bug does not occur. It does
not matter, whether the initrd contains and loads the module before
thawing.
So the fix is to invert the boolean parameter deep in all pm function
calls, which was clearly intended to be set like that.
First report was on Github [1], which you have to guess from the
resume logs in the posted dmesg snippet. Recently I posted one on
Bugzilla [2], since I did not have an AQC device so far.
#regzbot introduced: 8ce8427
#regzbot from: koo5 <[email protected]>
#regzbot monitor: Aquantia/AQtion#32
Fixes: 8aaa112 ("net: atlantic: refactoring pm logic")
Link: Aquantia/AQtion#32 [1]
Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2]
Cc: [email protected]
Reported-by: koo5 <[email protected]>
Signed-off-by: Manuel Ullmann <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>
tuxedo-bot
pushed a commit
to tuxedocomputers/linux
that referenced
this issue
Jun 23, 2022
BugLink: https://bugs.launchpad.net/bugs/1972905 commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]> Signed-off-by: Kamal Mostafa <[email protected]> Signed-off-by: Kleber Sacilotto de Souza <[email protected]>
it-is-a-robot
pushed a commit
to openeuler-mirror/kernel
that referenced
this issue
Jul 27, 2022
stable inclusion from stable-v5.10.113 commit ba2716da233618c3f361b8ece818a0e91de7a8f0 category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I5ISAH Reference: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=ba2716da233618c3f361b8ece818a0e91de7a8f0 -------------------------------- commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]> Signed-off-by: Zheng Zengkai <[email protected]> Acked-by: Xie XiuQi <[email protected]>
it-is-a-robot
pushed a commit
to openeuler-mirror/kernel
that referenced
this issue
Jul 27, 2022
stable inclusion from stable-v5.10.113 commit ba2716da233618c3f361b8ece818a0e91de7a8f0 category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I5ISAH Reference: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=ba2716da233618c3f361b8ece818a0e91de7a8f0 -------------------------------- commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]> Signed-off-by: Zheng Zengkai <[email protected]> Acked-by: Xie XiuQi <[email protected]>
it-is-a-robot
pushed a commit
to openeuler-mirror/kernel
that referenced
this issue
Jul 27, 2022
stable inclusion from stable-v5.10.113 commit ba2716da233618c3f361b8ece818a0e91de7a8f0 category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I5ISAH Reference: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=ba2716da233618c3f361b8ece818a0e91de7a8f0 -------------------------------- commit cbe6c3a upstream. This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]> Signed-off-by: Zheng Zengkai <[email protected]> Acked-by: Xie XiuQi <[email protected]>
andyhu-stf
pushed a commit
to starfive-tech/linux
that referenced
this issue
Feb 27, 2023
This will reset deeply on freeze and thaw instead of suspend and
resume and prevent null pointer dereferences of the uninitialized ring
0 buffer while thawing.
The impact is an indefinitely hanging kernel. You can't switch
consoles after this and the only possible user interaction is SysRq.
BUG: kernel NULL pointer dereference
RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic]
aq_vec_init+0x85/0xe0 [atlantic]
aq_nic_init+0xf7/0x1d0 [atlantic]
atl_resume_common+0x4f/0x100 [atlantic]
pci_pm_thaw+0x42/0xa0
resolves in aq_ring.o to
```
0000000000000ae0 <aq_ring_rx_fill>:
{
/* ... */
baf: 48 8b 43 08 mov 0x8(%rbx),%rax
buff->flags = 0U; /* buff is NULL */
```
The bug has been present since the introduction of the new pm code in
8aaa112 ("net: atlantic: refactoring pm logic") and was hidden
until 8ce8427 ("net: atlantic: changes for multi-TC support"),
which refactored the aq_vec_{free,alloc} functions into
aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The
original functions just always reinitialized the buffers on S3/S4. If
the interface is down before freezing, the bug does not occur. It does
not matter, whether the initrd contains and loads the module before
thawing.
So the fix is to invert the boolean parameter deep in all pm function
calls, which was clearly intended to be set like that.
First report was on Github [1], which you have to guess from the
resume logs in the posted dmesg snippet. Recently I posted one on
Bugzilla [2], since I did not have an AQC device so far.
#regzbot introduced: 8ce8427
#regzbot from: koo5 <[email protected]>
#regzbot monitor: Aquantia/AQtion#32
Fixes: 8aaa112 ("net: atlantic: refactoring pm logic")
Link: Aquantia/AQtion#32 [1]
Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2]
Cc: [email protected]
Reported-by: koo5 <[email protected]>
Signed-off-by: Manuel Ullmann <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
6543
pushed a commit
to 6543-forks/starfive_visionfive2_linux
that referenced
this issue
Feb 28, 2023
This will reset deeply on freeze and thaw instead of suspend and
resume and prevent null pointer dereferences of the uninitialized ring
0 buffer while thawing.
The impact is an indefinitely hanging kernel. You can't switch
consoles after this and the only possible user interaction is SysRq.
BUG: kernel NULL pointer dereference
RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic]
aq_vec_init+0x85/0xe0 [atlantic]
aq_nic_init+0xf7/0x1d0 [atlantic]
atl_resume_common+0x4f/0x100 [atlantic]
pci_pm_thaw+0x42/0xa0
resolves in aq_ring.o to
```
0000000000000ae0 <aq_ring_rx_fill>:
{
/* ... */
baf: 48 8b 43 08 mov 0x8(%rbx),%rax
buff->flags = 0U; /* buff is NULL */
```
The bug has been present since the introduction of the new pm code in
8aaa112 ("net: atlantic: refactoring pm logic") and was hidden
until 8ce8427 ("net: atlantic: changes for multi-TC support"),
which refactored the aq_vec_{free,alloc} functions into
aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The
original functions just always reinitialized the buffers on S3/S4. If
the interface is down before freezing, the bug does not occur. It does
not matter, whether the initrd contains and loads the module before
thawing.
So the fix is to invert the boolean parameter deep in all pm function
calls, which was clearly intended to be set like that.
First report was on Github [1], which you have to guess from the
resume logs in the posted dmesg snippet. Recently I posted one on
Bugzilla [2], since I did not have an AQC device so far.
#regzbot introduced: 8ce8427
#regzbot from: koo5 <[email protected]>
#regzbot monitor: Aquantia/AQtion#32
Fixes: 8aaa112 ("net: atlantic: refactoring pm logic")
Link: Aquantia/AQtion#32 [1]
Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2]
Cc: [email protected]
Reported-by: koo5 <[email protected]>
Signed-off-by: Manuel Ullmann <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
|
yup, not getting any errors in 5.15.0-75 |
PlaidCat
added a commit
to ctrliq/kernel-src-tree
that referenced
this issue
Sep 12, 2024
jira LE-1907 Rebuild_History Non-Buildable kernel-4.18.0-372.32.1.el8_6 commit-author Manuel Ullmann <[email protected]> commit cbe6c3a This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> (cherry picked from commit cbe6c3a) Signed-off-by: Jonathan Maple <[email protected]>
PlaidCat
added a commit
to ctrliq/kernel-src-tree
that referenced
this issue
Sep 13, 2024
jira LE-1907 Rebuild_History Non-Buildable kernel-4.18.0-448.el8 commit-author Manuel Ullmann <[email protected]> commit cbe6c3a This will reset deeply on freeze and thaw instead of suspend and resume and prevent null pointer dereferences of the uninitialized ring 0 buffer while thawing. The impact is an indefinitely hanging kernel. You can't switch consoles after this and the only possible user interaction is SysRq. BUG: kernel NULL pointer dereference RIP: 0010:aq_ring_rx_fill+0xcf/0x210 [atlantic] aq_vec_init+0x85/0xe0 [atlantic] aq_nic_init+0xf7/0x1d0 [atlantic] atl_resume_common+0x4f/0x100 [atlantic] pci_pm_thaw+0x42/0xa0 resolves in aq_ring.o to ``` 0000000000000ae0 <aq_ring_rx_fill>: { /* ... */ baf: 48 8b 43 08 mov 0x8(%rbx),%rax buff->flags = 0U; /* buff is NULL */ ``` The bug has been present since the introduction of the new pm code in 8aaa112 ("net: atlantic: refactoring pm logic") and was hidden until 8ce8427 ("net: atlantic: changes for multi-TC support"), which refactored the aq_vec_{free,alloc} functions into aq_vec_{,ring}_{free,alloc}, but is technically not wrong. The original functions just always reinitialized the buffers on S3/S4. If the interface is down before freezing, the bug does not occur. It does not matter, whether the initrd contains and loads the module before thawing. So the fix is to invert the boolean parameter deep in all pm function calls, which was clearly intended to be set like that. First report was on Github [1], which you have to guess from the resume logs in the posted dmesg snippet. Recently I posted one on Bugzilla [2], since I did not have an AQC device so far. #regzbot introduced: 8ce8427 #regzbot from: koo5 <[email protected]> #regzbot monitor: Aquantia/AQtion#32 Fixes: 8aaa112 ("net: atlantic: refactoring pm logic") Link: Aquantia/AQtion#32 [1] Link: https://bugzilla.kernel.org/show_bug.cgi?id=215798 [2] Cc: [email protected] Reported-by: koo5 <[email protected]> Signed-off-by: Manuel Ullmann <[email protected]> Signed-off-by: David S. Miller <[email protected]> (cherry picked from commit cbe6c3a) Signed-off-by: Jonathan Maple <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
The text was updated successfully, but these errors were encountered: