Re: Fwd: iwlmvm: Linux 6.7 pre-rc1 prints stack trace of iwl_op_mode_mvm_start; caused by commit b6e3d1ba4fcf

[Bagas Sanjaya]

Hi Niklāvs and Fabio,

On Fri, Nov 03, 2023 at 08:13:39PM +0700, Bagas Sanjaya wrote:
> Hi,
> 
> I notice a regression report on Bugzilla [1]. Quoting from it:
> 
> > When testing the branch that will become Linux 6.7-rc1, ever since commit b6e3d1ba4fcf02176846d03a930203d8133c0aaf I have observed the following abbreviated kernel stack trace (please see attachments for the full trace):
> > 
> > ```
> > [    6.733281]  ? __warn+0xc3/0x1d0
> > [    6.733284]  ? iwl_op_mode_mvm_start+0x9c4/0x9d0 [iwlmvm]
> > [    6.733299]  ? report_bug+0x167/0x220
> > [    6.733301]  ? handle_bug+0x3d/0x90
> > [    6.733304]  ? exc_invalid_op+0x1a/0x60
> > [    6.733306]  ? asm_exc_invalid_op+0x1a/0x20
> > [    6.733309]  ? iwl_op_mode_mvm_start+0x9c4/0x9d0 [iwlmvm]
> > [    6.733325]  _iwl_op_mode_start+0x66/0xe0 [iwlwifi]
> > [    6.733338]  iwl_opmode_register+0x87/0xd0 [iwlwifi]
> > [    6.733350]  ? __cfi_init_module+0x10/0x10 [iwlmvm]
> > [    6.733364]  init_module+0x22/0xff0 [iwlmvm]
> > [    6.733378]  ? __cfi_init_module+0x10/0x10 [iwlmvm]
> > [    6.733392]  do_one_initcall+0x129/0x380
> > [    6.733395]  ? idr_alloc_cyclic+0x148/0x1e0
> > [    6.733397]  ? security_kernfs_init_security+0x41/0x80
> > [    6.733399]  ? __kernfs_new_node+0x1be/0x250
> > [    6.733401]  ? preempt_count_add+0x55/0xb0
> > [    6.733404]  ? up_write+0x4a/0xe0
> > [    6.733406]  ? preempt_count_add+0x55/0xb0
> > [    6.733408]  ? sysvec_call_function+0xa4/0xb0
> > [    6.733410]  ? asm_sysvec_call_function+0x1a/0x20
> > [    6.733411]  ? free_unref_page_prepare+0xf3/0x410
> > [    6.733413]  ? preempt_count_add+0x62/0xb0
> > [    6.733415]  ? _raw_spin_trylock+0x19/0x60
> > [    6.733417]  ? _raw_spin_unlock+0x11/0x30
> > [    6.733419]  ? __kmem_cache_free+0x29a/0x3c0
> > [    6.733421]  ? vfree+0xd2/0x150
> > [    6.733422]  ? slab_post_alloc_hook+0x76/0x3d0
> > [    6.733425]  ? do_init_module+0x3f/0x230
> > [    6.733427]  ? __kmem_cache_alloc_node+0x1fe/0x2f0
> > [    6.733430]  do_init_module+0x7a/0x230
> > [    6.733432]  __se_sys_init_module+0x1a0/0x220
> > [    6.733435]  do_syscall_64+0x7a/0x100
> > [    6.733438]  ? syscall_exit_to_user_mode+0x2d/0x1d0
> > [    6.733440]  ? do_syscall_64+0x89/0x100
> > [    6.733442]  ? do_user_addr_fault+0x4e7/0x680
> > [    6.733444]  ? exc_page_fault+0x61/0x150
> > [    6.733446]  entry_SYSCALL_64_after_hwframe+0x6e/0x76
> > ```
> > 
> > I did try reverting the offending commit but the result did not compile and I did not want to dig deeper to identify any subsequent commits what would need reverting to fix compilation.
> > 
> > If it's relevant, I'm building the kernel with Clang 17.0.4 toolchain by setting the LLVM=1 variable. CFI is enabled but I'm not sure if it's actually functional.
> 
> See Bugzilla for the full thread.
> 
> Anyway, I'm adding this regression to regzbot:
> 
> #regzbot introduced: b6e3d1ba4fcf02 https://bugzilla.kernel.org/show_bug.cgi?id=218095
> #regzbot title: new iwlwifi firmware statistics API triggers stack trace
> 

Emmanuel had sent a proposed fix at [1]. Please test.

Thanks.

[1]: https://lore.kernel.org/regressions/20231112100944.21177-1-emmanuel.grumbach@xxxxxxxxx/

-- 
An old man doll... just what I always wanted! - Clara

Attachment: signature.asc
Description: PGP signature