Re: [PATCH] iwlwifi: mvm: protect regulatory_set_wiphy_regd_sync() with wiphy lock

[Coelho, Luciano]

On Wed, 2021-11-24 at 09:21 +0100, Jiri Kosina wrote:
> On Wed, 24 Nov 2021, Coelho, Luciano wrote:
> 
> > > > Since the switch away from rtnl to wiphy lock, 
> > > > regulatory_set_wiphy_regd_sync() has to be called with wiphy lock held; 
> > > > this is currently not the case on the module load codepath.
> > > > 
> > > > Fix that by properly acquiring it in iwl_mvm_start_get_nvm() to maintain 
> > > > also lock ordering against mvm->mutex and RTNL.
> > > > 
> > > > This fixes the splat below.
> > > > 
> > > >  =============================
> > > >  WARNING: suspicious RCU usage
> > > >  5.16.0-rc2 #1 Not tainted
> > > >  -----------------------------
> > > >  drivers/net/wireless/intel/iwlwifi/mvm/mac80211.c:264 suspicious rcu_dereference_protected() usage!
> > > > 
> > > >  other info that might help us debug this:
> > > > 
> > > >  rcu_scheduler_active = 2, debug_locks = 1
> > > >  3 locks held by modprobe/578:
> > > >   #0: ffffffffc0b6f0e8 (iwlwifi_opmode_table_mtx){+.+.}-{3:3}, at: iwl_opmode_register+0x2e/0xe0 [iwlwifi]
> > > >   #1: ffffffff9a856b08 (rtnl_mutex){+.+.}-{3:3}, at: iwl_op_mode_mvm_start+0xa0b/0xcb0 [iwlmvm]
> > > >   #2: ffff8e5242f53380 (&mvm->mutex){+.+.}-{3:3}, at: iwl_op_mode_mvm_start+0xa16/0xcb0 [iwlmvm]
> > > > 
> > > >  stack backtrace:
> > > >  CPU: 1 PID: 578 Comm: modprobe Not tainted 5.16.0-rc2 #1
> > > >  Hardware name: LENOVO 20K5S22R00/20K5S22R00, BIOS R0IET38W (1.16 ) 05/31/2017
> > > >  Call Trace:
> > > >   <TASK>
> > > >   dump_stack_lvl+0x58/0x71
> > > >   iwl_mvm_init_fw_regd+0x13d/0x180 [iwlmvm]
> > > >   iwl_mvm_init_mcc+0x66/0x1d0 [iwlmvm]
> > > >   iwl_op_mode_mvm_start+0xc6d/0xcb0 [iwlmvm]
> > > >   _iwl_op_mode_start.isra.4+0x42/0x80 [iwlwifi]
> > > >   iwl_opmode_register+0x71/0xe0 [iwlwifi]
> > > >   ? 0xffffffffc1062000
> > > >   iwl_mvm_init+0x34/0x1000 [iwlmvm]
> > > >   do_one_initcall+0x5b/0x300
> > > >   do_init_module+0x5b/0x21c
> > > >   load_module+0x1b2f/0x2320
> > > >   ? __do_sys_finit_module+0xaa/0x110
> > > >   __do_sys_finit_module+0xaa/0x110
> > > >   do_syscall_64+0x3a/0xb0
> > > >   entry_SYSCALL_64_after_hwframe+0x44/0xae
> > > >  RIP: 0033:0x7f7cdd7c8ded
> > > >  Code: 5b 41 5c c3 66 0f 1f 84 00 00 00 00 00 f3 0f 1e fa 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d fb ef 0e 00 f7 d8 64 89 01 48
> > > >  RSP: 002b:00007fffb90bf458 EFLAGS: 00000246 ORIG_RAX: 0000000000000139
> > > >  RAX: ffffffffffffffda RBX: 0000559c501caf00 RCX: 00007f7cdd7c8ded
> > > >  RDX: 0000000000000000 RSI: 0000559c4eb366ee RDI: 0000000000000002
> > > >  RBP: 0000000000040000 R08: 0000000000000000 R09: 0000559c501ca9f8
> > > >  R10: 0000000000000002 R11: 0000000000000246 R12: 0000559c4eb366ee
> > > >  R13: 0000559c501cadb0 R14: 0000000000000000 R15: 0000559c501cbad0
> > > >   </TASK>
> > > >  ------------[ cut here ]------------
> > > >  WARNING: CPU: 1 PID: 578 at net/wireless/reg.c:3107 reg_process_self_managed_hint+0x183/0x1d0 [cfg80211]
> > > >  Modules linked in:
> > > >  CPU: 1 PID: 578 Comm: modprobe Not tainted 5.16.0-rc2 #1
> > > >  Hardware name: LENOVO 20K5S22R00/20K5S22R00, BIOS R0IET38W (1.16 ) 05/31/2017
> > > >  RIP: 0010:reg_process_self_managed_hint+0x183/0x1d0 [cfg80211]
> > > >  Code: 83 c4 60 5b 41 5a 41 5c 41 5d 41 5e 41 5f 5d 49 8d 62 f8 c3 48 8d 7b 68 be ff ff ff ff e8 75 4a 13 d9 85 c0 0f 85 e6 fe ff ff <0f> 0b e9 df fe ff ff 0f 0b 80 3d bc 2c 0b 00 00 0f 85 c2 fe ff ff
> > > >  RSP: 0018:ffff9994809cfaf0 EFLAGS: 00010246
> > > >  RAX: 0000000000000000 RBX: ffff8e5242f505c0 RCX: 0000000000000000
> > > >  RDX: 0000000000000000 RSI: ffff8e5242f50628 RDI: ffff8e524a2b5cd0
> > > >  RBP: ffff9994809cfb80 R08: 0000000000000001 R09: ffffffff9b2e2f50
> > > >  R10: ffff9994809cfb98 R11: ffffffffffffffff R12: 0000000000000000
> > > >  R13: ffff8e5242f532e8 R14: ffff8e5248914010 R15: ffff8e5242f532e0
> > > >  FS:  00007f7cdd6af740(0000) GS:ffff8e5367480000(0000) knlGS:0000000000000000
> > > >  CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> > > >  CR2: 00007f19430687ac CR3: 00000001088fa003 CR4: 00000000003706e0
> > > >  Call Trace:
> > > >   <TASK>
> > > >   ? lock_is_held_type+0xb4/0x120
> > > >   ? regulatory_set_wiphy_regd_sync+0x2f/0x80 [cfg80211]
> > > >   regulatory_set_wiphy_regd_sync+0x2f/0x80 [cfg80211]
> > > >   iwl_mvm_init_mcc+0xcd/0x1d0 [iwlmvm]
> > > >   iwl_op_mode_mvm_start+0xc6d/0xcb0 [iwlmvm]
> > > >   _iwl_op_mode_start.isra.4+0x42/0x80 [iwlwifi]
> > > >   iwl_opmode_register+0x71/0xe0 [iwlwifi]
> > > >   ? 0xffffffffc1062000
> > > >   iwl_mvm_init+0x34/0x1000 [iwlmvm]
> > > >   do_one_initcall+0x5b/0x300
> > > >   do_init_module+0x5b/0x21c
> > > >   load_module+0x1b2f/0x2320
> > > >   ? __do_sys_finit_module+0xaa/0x110
> > > >   __do_sys_finit_module+0xaa/0x110
> > > >   do_syscall_64+0x3a/0xb0
> > > >   entry_SYSCALL_64_after_hwframe+0x44/0xae
> > > >  RIP: 0033:0x7f7cdd7c8ded
> > > >  Code: 5b 41 5c c3 66 0f 1f 84 00 00 00 00 00 f3 0f 1e fa 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d fb ef 0e 00 f7 d8 64 89 01 48
> > > >  RSP: 002b:00007fffb90bf458 EFLAGS: 00000246 ORIG_RAX: 0000000000000139
> > > >  RAX: ffffffffffffffda RBX: 0000559c501caf00 RCX: 00007f7cdd7c8ded
> > > >  RDX: 0000000000000000 RSI: 0000559c4eb366ee RDI: 0000000000000002
> > > >  RBP: 0000000000040000 R08: 0000000000000000 R09: 0000559c501ca9f8
> > > >  R10: 0000000000000002 R11: 0000000000000246 R12: 0000559c4eb366ee
> > > >  R13: 0000559c501cadb0 R14: 0000000000000000 R15: 0000559c501cbad0
> > > > 
> > > > Fixes: a05829a7222e9d1 ("cfg80211: avoid holding the RTNL when calling the driver")
> > > > Signed-off-by: Jiri Kosina <jkosina@xxxxxxx>
> > > 
> > > I think this should go to wireless-drivers so I assigned this to me.
> > > Luca, ack?
> > 
> > Hmmm, I thought we already had a fix for this as well?
> 
> FWIW I don't see this fixed in latest linux-next.

You're right, it's not there.  For some reason that is in our internal
tree and came from a merge from upstream, which is strange.

But regardless of the reason why it's not there, it should be applied,
so thanks! :)

Acked-by: Luca Coelho <luciano.coelho@xxxxxxxxx>

--
Cheers,
Luca.