Slab bug during skb allocation

Slab bug during skb allocation

[Md Muazzam Husain]

[-- Attachment #1: Type: text/plain, Size: 1171 bytes --]

Hi All,
         I have kernel version 3.10.59. So during the allocation for skb i
am getting call trace for __netdev_alloc_skb which follows
till cache_alloc_refill and finally crashes. Please find snapshot below.
addr2line -e vmlinux ffffffffc114109c shows linux/mm/slab.c:2700 has the
problem. Is there any existing bug and any patch for this.

Call Trace:
[<ffffffffc114109c>] *cache_alloc_refill*+0x1b4/0x848
[<ffffffffc1141854>] __kmalloc_track_caller+0x124/0x1f0
[<ffffffffc144a930>] __kmalloc_reserve.isra.17+0x40/0xa8
[<ffffffffc144b644>] __alloc_skb+0x84/0x1a0
[<ffffffffc144d084>] *__netdev_alloc_skb*+0xdc/0x148
[<ffffffffc13fb510>] xlpnae_irq+0x148/0x200
[<ffffffffc10b87ac>] handle_irq_event_percpu+0x94/0x3f0
[<ffffffffc10b8b5c>] handle_irq_event+0x54/0x98
[<ffffffffc10bc724>] handle_level_irq+0xe4/0x170
[<ffffffffc10b7d90>] generic_handle_irq+0x38/0x58
[<ffffffffc1018a20>] do_IRQ+0x18/0x28
[<ffffffffc1015ce8>] plat_irq_dispatch+0x258/0x490
[<ffffffffc1016a00>] ret_from_irq+0x0/0x4
[<ffffffffc1016ca0>] __r4k_wait+0x20/0x40
[<ffffffffc109072c>] cpu_startup_entry+0xd4/0x2d8
[<ffffffffc1014ab8>] prom_pre_boot_secondary_cpus+0x0/0x48


Thanks,
Muazzam

[-- Attachment #2: Type: text/html, Size: 1485 bytes --]

Re: Slab bug during skb allocation

[Matthew Wilcox]

On Thu, Dec 10, 2020 at 11:25:13PM +0530, Md Muazzam Husain wrote:
> Hi All,
>          I have kernel version 3.10.59. So during the allocation for skb i

That kernel is over 6 years old: https://lwn.net/Articles/618650/ It's
been out of support since November 2017.  The development kernel it was
based on is over 7 years old.  It's unreasonable to expect people to
know the answer to this.

Can you reproduce this problem with a recent kernel?

> am getting call trace for __netdev_alloc_skb which follows
> till cache_alloc_refill and finally crashes. Please find snapshot below.
> addr2line -e vmlinux ffffffffc114109c shows linux/mm/slab.c:2700 has the
> problem. Is there any existing bug and any patch for this.
> 
> Call Trace:
> [<ffffffffc114109c>] *cache_alloc_refill*+0x1b4/0x848
> [<ffffffffc1141854>] __kmalloc_track_caller+0x124/0x1f0
> [<ffffffffc144a930>] __kmalloc_reserve.isra.17+0x40/0xa8
> [<ffffffffc144b644>] __alloc_skb+0x84/0x1a0
> [<ffffffffc144d084>] *__netdev_alloc_skb*+0xdc/0x148
> [<ffffffffc13fb510>] xlpnae_irq+0x148/0x200
> [<ffffffffc10b87ac>] handle_irq_event_percpu+0x94/0x3f0
> [<ffffffffc10b8b5c>] handle_irq_event+0x54/0x98
> [<ffffffffc10bc724>] handle_level_irq+0xe4/0x170
> [<ffffffffc10b7d90>] generic_handle_irq+0x38/0x58
> [<ffffffffc1018a20>] do_IRQ+0x18/0x28
> [<ffffffffc1015ce8>] plat_irq_dispatch+0x258/0x490
> [<ffffffffc1016a00>] ret_from_irq+0x0/0x4
> [<ffffffffc1016ca0>] __r4k_wait+0x20/0x40
> [<ffffffffc109072c>] cpu_startup_entry+0xd4/0x2d8
> [<ffffffffc1014ab8>] prom_pre_boot_secondary_cpus+0x0/0x48
> 
> 
> Thanks,
> Muazzam

Re: Slab bug during skb allocation

[Md Muazzam Husain]

[-- Attachment #1: Type: text/plain, Size: 1994 bytes --]

Thanks for the information.
Actually this issue is not coming for lab units, but under high traffic
where irqs are hit frequently it's coming. Actually i have used GFP_NOWAIT
flag, can i use GFP_ATOMIC(Since it uses some emergency pool).

Thanks,
Muazzam


On Thu, Dec 10, 2020 at 11:42 PM Matthew Wilcox <willy@infradead.org> wrote:

> On Thu, Dec 10, 2020 at 11:25:13PM +0530, Md Muazzam Husain wrote:
> > Hi All,
> >          I have kernel version 3.10.59. So during the allocation for skb
> i
>
> That kernel is over 6 years old: https://lwn.net/Articles/618650/ It's
> been out of support since November 2017.  The development kernel it was
> based on is over 7 years old.  It's unreasonable to expect people to
> know the answer to this.
>
> Can you reproduce this problem with a recent kernel?
>
> > am getting call trace for __netdev_alloc_skb which follows
> > till cache_alloc_refill and finally crashes. Please find snapshot below.
> > addr2line -e vmlinux ffffffffc114109c shows linux/mm/slab.c:2700 has the
> > problem. Is there any existing bug and any patch for this.
> >
> > Call Trace:
> > [<ffffffffc114109c>] *cache_alloc_refill*+0x1b4/0x848
> > [<ffffffffc1141854>] __kmalloc_track_caller+0x124/0x1f0
> > [<ffffffffc144a930>] __kmalloc_reserve.isra.17+0x40/0xa8
> > [<ffffffffc144b644>] __alloc_skb+0x84/0x1a0
> > [<ffffffffc144d084>] *__netdev_alloc_skb*+0xdc/0x148
> > [<ffffffffc13fb510>] xlpnae_irq+0x148/0x200
> > [<ffffffffc10b87ac>] handle_irq_event_percpu+0x94/0x3f0
> > [<ffffffffc10b8b5c>] handle_irq_event+0x54/0x98
> > [<ffffffffc10bc724>] handle_level_irq+0xe4/0x170
> > [<ffffffffc10b7d90>] generic_handle_irq+0x38/0x58
> > [<ffffffffc1018a20>] do_IRQ+0x18/0x28
> > [<ffffffffc1015ce8>] plat_irq_dispatch+0x258/0x490
> > [<ffffffffc1016a00>] ret_from_irq+0x0/0x4
> > [<ffffffffc1016ca0>] __r4k_wait+0x20/0x40
> > [<ffffffffc109072c>] cpu_startup_entry+0xd4/0x2d8
> > [<ffffffffc1014ab8>] prom_pre_boot_secondary_cpus+0x0/0x48
> >
> >
> > Thanks,
> > Muazzam
>

[-- Attachment #2: Type: text/html, Size: 2729 bytes --]