virtio_console: allocate the port_buffer with the caller's gfp

[PATCH] virtio_console: allocate the port_buffer with the caller's gfp

Posted by Breno Leitao 1 month, 2 weeks ago

put_chars() runs from the hvc console write path with preemption
disabled, so it asks alloc_buf() for GFP_ATOMIC. Only the data buffer
gets it: the struct port_buffer itself keeps the GFP_KERNEL default, so
the allocation can enter direct reclaim and sleep. A write to /dev/kmsg
on a CONFIG_DEBUG_ATOMIC_SLEEP kernel splats:

  BUG: sleeping function called from invalid context at ./include/linux/sched/mm.h:320
  in_atomic(): 1, irqs_disabled(): 1, non_block: 0, pid: 1, name: virtme-ng-init
  preempt_count: 1, expected: 0
  Preemption disabled at:
  [<ffffffff813fd90d>] vprintk_emit+0x17d/0x510
  Call Trace:
   <TASK>
   dump_stack_lvl+0x69/0xa0
   __might_resched+0x37a/0x4d0
   __kmalloc_cache_noprof+0x94/0x5f0
   put_chars+0x209/0x3e0
   hvc_console_print+0x234/0x640
   console_flush_all+0x4fc/0x950
   console_unlock+0xbf/0x1b0
   vprintk_emit+0x312/0x510
   devkmsg_emit+0xba/0x110
   devkmsg_write+0x21b/0x2e0
   vfs_write+0x4dc/0x9d0
   ksys_write+0x108/0x1e0
   do_syscall_64+0xfa/0x460
   </TASK>

Pass gfp on to that allocation too.

Fixes: fc220d6be3c7 ("virtio_console: refactor __send_to_port() buffer ownership")
Signed-off-by: Breno Leitao <leitao@debian.org>
---
 drivers/char/virtio_console.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/char/virtio_console.c b/drivers/char/virtio_console.c
index 62eecfa61646d..7f6cbe851d1e1 100644
--- a/drivers/char/virtio_console.c
+++ b/drivers/char/virtio_console.c
@@ -426,7 +426,7 @@ static struct port_buffer *alloc_buf(struct virtio_device *vdev, size_t buf_size
 	 * Allocate buffer and the sg list. The sg list array is allocated
 	 * directly after the port_buffer struct.
 	 */
-	buf = kmalloc_flex(*buf, sg, pages);
+	buf = kmalloc_flex(*buf, sg, pages, gfp);
 	if (!buf)
 		goto fail;
 

---
base-commit: 6b8c8af514d739d0335f5579b585e02babe8a727
change-id: 20260810-serial-d9b259ec9100

Best regards,
--  
Breno Leitao <leitao@debian.org>

Re: [PATCH] virtio_console: allocate the port_buffer with the caller's gfp

Posted by Sungho Bae 1 month, 1 week ago

Hi Breno,

(Sorry, sending this again as I forgot to use reply-all in my previous email.)

On Tue, Aug 11, 2026 at 1:46 AM Breno Leitao <leitao@debian.org> wrote:
>
> put_chars() runs from the hvc console write path with preemption
> disabled, so it asks alloc_buf() for GFP_ATOMIC. Only the data buffer
> gets it: the struct port_buffer itself keeps the GFP_KERNEL default, so
> the allocation can enter direct reclaim and sleep. A write to /dev/kmsg
> on a CONFIG_DEBUG_ATOMIC_SLEEP kernel splats:
>
>   BUG: sleeping function called from invalid context at ./include/linux/sched/mm.h:320
>   in_atomic(): 1, irqs_disabled(): 1, non_block: 0, pid: 1, name: virtme-ng-init
>   preempt_count: 1, expected: 0
>   Preemption disabled at:
>   [<ffffffff813fd90d>] vprintk_emit+0x17d/0x510
>   Call Trace:
>    <TASK>
>    dump_stack_lvl+0x69/0xa0
>    __might_resched+0x37a/0x4d0
>    __kmalloc_cache_noprof+0x94/0x5f0
>    put_chars+0x209/0x3e0
>    hvc_console_print+0x234/0x640
>    console_flush_all+0x4fc/0x950
>    console_unlock+0xbf/0x1b0
>    vprintk_emit+0x312/0x510
>    devkmsg_emit+0xba/0x110
>    devkmsg_write+0x21b/0x2e0
>    vfs_write+0x4dc/0x9d0
>    ksys_write+0x108/0x1e0
>    do_syscall_64+0xfa/0x460
>    </TASK>
>
> Pass gfp on to that allocation too.
>
> Fixes: fc220d6be3c7 ("virtio_console: refactor __send_to_port() buffer ownership")
> Signed-off-by: Breno Leitao <leitao@debian.org>


Looks good to me.
Thanks for fixing the part I missed.

Acked-by: Sungho Bae <baver.bae@lge.com>

--
Thanks,
Sungho

Re: [PATCH] virtio_console: allocate the port_buffer with the caller's gfp

Posted by Breno Leitao 1 month ago

On Wed, Aug 19, 2026 at 08:36:25PM +0900, Sungho Bae wrote:
> >
> > put_chars() runs from the hvc console write path with preemption
> > disabled, so it asks alloc_buf() for GFP_ATOMIC. Only the data buffer
> > gets it: the struct port_buffer itself keeps the GFP_KERNEL default, so
> > the allocation can enter direct reclaim and sleep. A write to /dev/kmsg
> > on a CONFIG_DEBUG_ATOMIC_SLEEP kernel splats:
> >
> >   BUG: sleeping function called from invalid context at ./include/linux/sched/mm.h:320
> >   in_atomic(): 1, irqs_disabled(): 1, non_block: 0, pid: 1, name: virtme-ng-init
> >   preempt_count: 1, expected: 0
> >   Preemption disabled at:
> >   [<ffffffff813fd90d>] vprintk_emit+0x17d/0x510
> >   Call Trace:
> >    <TASK>
> >    dump_stack_lvl+0x69/0xa0
> >    __might_resched+0x37a/0x4d0
> >    __kmalloc_cache_noprof+0x94/0x5f0
> >    put_chars+0x209/0x3e0
> >    hvc_console_print+0x234/0x640
> >    console_flush_all+0x4fc/0x950
> >    console_unlock+0xbf/0x1b0
> >    vprintk_emit+0x312/0x510
> >    devkmsg_emit+0xba/0x110
> >    devkmsg_write+0x21b/0x2e0
> >    vfs_write+0x4dc/0x9d0
> >    ksys_write+0x108/0x1e0
> >    do_syscall_64+0xfa/0x460
> >    </TASK>
> >
> > Pass gfp on to that allocation too.
> >
> > Fixes: fc220d6be3c7 ("virtio_console: refactor __send_to_port() buffer ownership")
> > Signed-off-by: Breno Leitao <leitao@debian.org>
> 
> Acked-by: Sungho Bae <baver.bae@lge.com>

Greg, are you the one that should pick this up?

The regression fixed here is now on Linus' tree.

Thanks
--breno

Re: [PATCH] virtio_console: allocate the port_buffer with the caller's gfp

Posted by Greg KH 1 month ago

On Fri, Aug 28, 2026 at 10:21:50AM -0700, Breno Leitao wrote:
> On Wed, Aug 19, 2026 at 08:36:25PM +0900, Sungho Bae wrote:
> > >
> > > put_chars() runs from the hvc console write path with preemption
> > > disabled, so it asks alloc_buf() for GFP_ATOMIC. Only the data buffer
> > > gets it: the struct port_buffer itself keeps the GFP_KERNEL default, so
> > > the allocation can enter direct reclaim and sleep. A write to /dev/kmsg
> > > on a CONFIG_DEBUG_ATOMIC_SLEEP kernel splats:
> > >
> > >   BUG: sleeping function called from invalid context at ./include/linux/sched/mm.h:320
> > >   in_atomic(): 1, irqs_disabled(): 1, non_block: 0, pid: 1, name: virtme-ng-init
> > >   preempt_count: 1, expected: 0
> > >   Preemption disabled at:
> > >   [<ffffffff813fd90d>] vprintk_emit+0x17d/0x510
> > >   Call Trace:
> > >    <TASK>
> > >    dump_stack_lvl+0x69/0xa0
> > >    __might_resched+0x37a/0x4d0
> > >    __kmalloc_cache_noprof+0x94/0x5f0
> > >    put_chars+0x209/0x3e0
> > >    hvc_console_print+0x234/0x640
> > >    console_flush_all+0x4fc/0x950
> > >    console_unlock+0xbf/0x1b0
> > >    vprintk_emit+0x312/0x510
> > >    devkmsg_emit+0xba/0x110
> > >    devkmsg_write+0x21b/0x2e0
> > >    vfs_write+0x4dc/0x9d0
> > >    ksys_write+0x108/0x1e0
> > >    do_syscall_64+0xfa/0x460
> > >    </TASK>
> > >
> > > Pass gfp on to that allocation too.
> > >
> > > Fixes: fc220d6be3c7 ("virtio_console: refactor __send_to_port() buffer ownership")
> > > Signed-off-by: Breno Leitao <leitao@debian.org>
> > 
> > Acked-by: Sungho Bae <baver.bae@lge.com>
> 
> Greg, are you the one that should pick this up?
> 
> The regression fixed here is now on Linus' tree.

I can, after -rc1 is out, sorry.

greg k-h

Re: [PATCH] virtio_console: allocate the port_buffer with the caller's gfp

Posted by Breno Leitao 1 month, 1 week ago

On Mon, Aug 10, 2026 at 09:40:00AM -0700, Breno Leitao wrote:
> put_chars() runs from the hvc console write path with preemption
> disabled, so it asks alloc_buf() for GFP_ATOMIC. Only the data buffer
> gets it: the struct port_buffer itself keeps the GFP_KERNEL default, so
> the allocation can enter direct reclaim and sleep. A write to /dev/kmsg
> on a CONFIG_DEBUG_ATOMIC_SLEEP kernel splats:
> 
>   BUG: sleeping function called from invalid context at ./include/linux/sched/mm.h:320
>   in_atomic(): 1, irqs_disabled(): 1, non_block: 0, pid: 1, name: virtme-ng-init
>   preempt_count: 1, expected: 0
>   Preemption disabled at:
>   [<ffffffff813fd90d>] vprintk_emit+0x17d/0x510
>   Call Trace:
>    <TASK>
>    dump_stack_lvl+0x69/0xa0
>    __might_resched+0x37a/0x4d0
>    __kmalloc_cache_noprof+0x94/0x5f0
>    put_chars+0x209/0x3e0
>    hvc_console_print+0x234/0x640
>    console_flush_all+0x4fc/0x950
>    console_unlock+0xbf/0x1b0
>    vprintk_emit+0x312/0x510
>    devkmsg_emit+0xba/0x110
>    devkmsg_write+0x21b/0x2e0
>    vfs_write+0x4dc/0x9d0
>    ksys_write+0x108/0x1e0
>    do_syscall_64+0xfa/0x460
>    </TASK>
> 
> Pass gfp on to that allocation too.
> 
> Fixes: fc220d6be3c7 ("virtio_console: refactor __send_to_port() buffer ownership")
> Signed-off-by: Breno Leitao <leitao@debian.org>

Ping?

Re: [PATCH] virtio_console: allocate the port_buffer with the caller's gfp

Posted by Sungho Bae 1 month, 1 week ago

I am sorry for the late reply.
I fully agree with your patch.
It was my mistake. I should have to allocate the buffer with gfp flags.
Thank you for the corrections.


On 2026년 8월 18일 오후 6시 6분 37초 GMT+09:00, Breno Leitao <leitao@debian.org> 작성함:
>On Mon, Aug 10, 2026 at 09:40:00AM -0700, Breno Leitao wrote:
>> put_chars() runs from the hvc console write path with preemption
>> disabled, so it asks alloc_buf() for GFP_ATOMIC. Only the data buffer
>> gets it: the struct port_buffer itself keeps the GFP_KERNEL default, so
>> the allocation can enter direct reclaim and sleep. A write to /dev/kmsg
>> on a CONFIG_DEBUG_ATOMIC_SLEEP kernel splats:
>> 
>>   BUG: sleeping function called from invalid context at ./include/linux/sched/mm.h:320
>>   in_atomic(): 1, irqs_disabled(): 1, non_block: 0, pid: 1, name: virtme-ng-init
>>   preempt_count: 1, expected: 0
>>   Preemption disabled at:
>>   [<ffffffff813fd90d>] vprintk_emit+0x17d/0x510
>>   Call Trace:
>>    <TASK>
>>    dump_stack_lvl+0x69/0xa0
>>    __might_resched+0x37a/0x4d0
>>    __kmalloc_cache_noprof+0x94/0x5f0
>>    put_chars+0x209/0x3e0
>>    hvc_console_print+0x234/0x640
>>    console_flush_all+0x4fc/0x950
>>    console_unlock+0xbf/0x1b0
>>    vprintk_emit+0x312/0x510
>>    devkmsg_emit+0xba/0x110
>>    devkmsg_write+0x21b/0x2e0
>>    vfs_write+0x4dc/0x9d0
>>    ksys_write+0x108/0x1e0
>>    do_syscall_64+0xfa/0x460
>>    </TASK>
>> 
>> Pass gfp on to that allocation too.
>> 
>> Fixes: fc220d6be3c7 ("virtio_console: refactor __send_to_port() buffer ownership")
>> Signed-off-by: Breno Leitao <leitao@debian.org>
>
>Ping? 
>

Re: [PATCH] virtio_console: allocate the port_buffer with the caller's gfp

Posted by Breno Leitao 1 month, 1 week ago

Hello Sungho,

On Wed, Aug 19, 2026 at 11:39:26AM +0900, Sungho Bae wrote:
> I am sorry for the late reply.
> I fully agree with your patch.
> It was my mistake. I should have to allocate the buffer with gfp flags.
> Thank you for the corrections.

Thanks, would you mind adding your reviewed-by or acked-by tag, please?

--breno