Re: [RFC PATCH] x86/apic: Fix BUG due to multiple allocation of legacy vectors.

[imran . f . khan]

On 20/5/21 3:56 pm, Greg KH wrote:
> On Wed, May 19, 2021 at 11:39:28PM +0000, Imran Khan wrote:
> > During activation of secondary CPUs, lapic_online is
> > invoked to initialize vectors. While lapic_online
> > installs legacy vectors on all CPUs, it does not set
> > the corresponding bits in per CPU bitmap maintained
> > under irq_matrix.
> > This may result in these legacy vectors getting allocated
> > by irq_matrix_alloc and if that happens subsequent invocation
> > of apic_update_vector will cause BUG like the one shown below:
> >
> > [  154.738226] kernel BUG at arch/x86/kernel/apic/vector.c:172!
> > [  154.805956] invalid opcode: 0000 [#1] SMP PTI
> > [  154.858092] CPU: 22 PID: 3569 Comm: ifup-eth Not tainted 5.8.0-20200716.x86_64 #1
> > [  154.954939] Hardware name: Oracle Corporation ORACLE SERVER X6-2/ASM,MOTHERBOARD,1U
> > [  155.073636] RIP: 0010:apic_update_vector+0xa7/0x190
> > [  155.131996] Code: 01 00 4a 8b 14 ed 80 69 01 a6 48 89 c8 4a 8d 04 e0 48 8b 04 10 48
> > 85 c0 0f 84 d2 00 00 00 48 3d 00 f0 ff ff 0f 87 c6 00 00 00 <0f> 0b 41 8b 46 10 48 0f
[.....]
> > [  156.268168] Call Trace:
> > [  156.297409]  ? irq_matrix_alloc+0x8a/0x150
> > [  156.346408]  assign_vector_locked+0xd2/0x170
> > [  156.397489]  x86_vector_activate+0x1b5/0x320
> > [  156.448570]  __irq_domain_activate_irq+0x64/0xa0
> > [  156.503808]  __irq_domain_activate_irq+0x38/0xa0
> > [  156.559050]  irq_domain_activate_irq+0x2b/0x40
> > [  156.612213]  irq_activate+0x25/0x30
> > [  156.653930]  __setup_irq+0x58f/0x7b0
> > [  156.696690]  request_threaded_irq+0xf8/0x1b0
> > [  156.747784]  ixgbe_open+0x3af/0x600 [ixgbe]
> > [  156.797827]  __dev_open+0xd8/0x160
> > [  156.838503]  dev_open+0x48/0x90
> > [  156.876065]  bond_enslave+0x2b6/0x12c0 [bonding]
> > [  156.931310]  ? vsscanf+0x5af/0x8e0
> > [  156.971986]  ? sscanf+0x4e/0x70
> > [  157.009546]  bond_option_slaves_set+0x112/0x1c0 [bonding]
> > [  157.074148]  __bond_opt_set+0xdc/0x320 [bonding]
> > [  157.129389]  __bond_opt_set_notify+0x2c/0x90 [bonding]
> > [  157.190871]  bond_opt_tryset_rtnl+0x56/0xa0 [bonding]
> > [  157.251315]  bonding_sysfs_store_option+0x52/0x90 [bonding]
> >
> > This patch marks these legacy vectors as assigned in irq_matrix
> > so that corresponding bits in percpu bitmaps get set and these
> > legacy vectors don't get reallocted.
> >
> > Signed-off-by: Imran Khan <imran.f.khan@xxxxxxxxxx>
> > ---
> >   arch/x86/kernel/apic/vector.c | 14 ++++++++++++--
> >   1 file changed, 12 insertions(+), 2 deletions(-)
>
> <formletter>
>
> This is not the correct way to submit patches for inclusion in the
> stable kernel tree.  Please read:
>      https://urldefense.com/v3/__https://www.kernel.org/doc/html/latest/process/stable-kernel-rules.html__;!!GqivPVa7Brio!MYRIHQM1qBB0Raf823KeG1-OSUDCwlOyOxqTp5fzHTlxAL1H4LZW6XniBtajqKVb3w$
> for how to do this properly.
>
> </formletter>

Thanks for clarifying the process and providing the relevant doc. Looks 
like CC-ing stable list for RFC patch was a mistake in first place. I 
will wait for review comments and, if the patch gets included, will 
inform via option 2 of the above mentioned document.

Thanks,
Imran