wifi: brcmfmac: prevent double-free on hardware-reset
authorDanny van Heumen <danny@dannyvanheumen.nl>
Mon, 11 Jul 2022 23:21:16 +0000 (23:21 +0000)
committerKalle Valo <kvalo@kernel.org>
Thu, 28 Jul 2022 09:59:05 +0000 (12:59 +0300)
commitcb774bd35318c1b4cb61f6f2caac85537d07fbde
tree5421531a77c2a70fa03e4255ec1b6b2138c76475
parent5c54ab24377b999c7f1c30b41218e6490cd4ac80
wifi: brcmfmac: prevent double-free on hardware-reset

In case of buggy firmware, brcmfmac may perform a hardware reset. If during
reset and subsequent probing an early failure occurs, a memory region is
accidentally double-freed. With hardened memory allocation enabled, this error
will be detected.

- return early where appropriate to skip unnecessary clean-up.
- set '.freezer' pointer to NULL to prevent double-freeing under possible
  other circumstances and to re-align result under various different
  behaviors of memory allocation freeing.
- correctly claim host on func1 for disabling func2.
- after reset, do not initiate probing immediately, but rely on events.

Given a firmware crash, function 'brcmf_sdio_bus_reset' is called. It calls
'brcmf_sdiod_remove', then follows up with 'brcmf_sdiod_probe' to reinitialize
the hardware. If 'brcmf_sdiod_probe' fails to "set F1 blocksize", it exits
early, which includes calling 'brcmf_sdiod_remove'. In both cases
'brcmf_sdiod_freezer_detach' is called to free allocated '.freezer', which
has not yet been re-allocated the second time.

Stacktrace of (failing) hardware reset after firmware-crash:

Code: b9402b82 8b0202c0 eb1a02df 54000041 (d4210000)
 ret_from_fork+0x10/0x20
 kthread+0x154/0x160
 worker_thread+0x188/0x504
 process_one_work+0x1f4/0x490
 brcmf_core_bus_reset+0x34/0x44 [brcmfmac]
 brcmf_sdio_bus_reset+0x68/0xc0 [brcmfmac]
 brcmf_sdiod_probe+0x170/0x21c [brcmfmac]
 brcmf_sdiod_remove+0x48/0xc0 [brcmfmac]
 kfree+0x210/0x220
 __slab_free+0x58/0x40c
Call trace:
x2 : 0000000000000040 x1 : fffffc00002d2b80 x0 : ffff00000b4aee40
x5 : ffff8000013fa728 x4 : 0000000000000001 x3 : ffff00000b4aee00
x8 : ffff800009967ce0 x7 : ffff8000099bfce0 x6 : 00000006f8005d01
x11: ffff8000099bfce0 x10: 00000000fffff000 x9 : ffff8000083401d0
x14: 0000000000000000 x13: 657a69736b636f6c x12: 6220314620746573
x17: 0000000000000000 x16: 0000000000000000 x15: 0000000000000030
x20: fffffc00002d2ba0 x19: fffffc00002d2b80 x18: 0000000000000000
x23: ffff00000b4aee00 x22: ffff00000b4aee00 x21: 0000000000000001
x26: ffff00000b4aee00 x25: ffff0000f7753705 x24: 000000000001288a
x29: ffff80000a22bbf0 x28: ffff000000401200 x27: 000000008020001a
sp : ffff80000a22bbf0
lr : kfree+0x210/0x220
pc : __slab_free+0x58/0x40c
pstate: 60000005 (nZCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
Workqueue: events brcmf_core_bus_reset [brcmfmac]
Hardware name: Pine64 Pinebook Pro (DT)
CPU: 2 PID: 639 Comm: kworker/2:2 Tainted: G         C        5.16.0-0.bpo.4-arm64 #1  Debian 5.16.12-1~bpo11+1
 nvmem_rockchip_efuse industrialio_triggered_buffer videodev snd_soc_core snd_pcm_dmaengine kfifo_buf snd_pcm io_domain mc industrialio mt>
Modules linked in: snd_seq_dummy snd_hrtimer snd_seq snd_seq_device nft_fib_inet nft_fib_ipv4 nft_fib_ipv6 nft_fib nft_reject_inet nf_reje>
Internal error: Oops - BUG: 0 [#1] SMP
kernel BUG at mm/slub.c:379!

Signed-off-by: Danny van Heumen <danny@dannyvanheumen.nl>
Reviewed-by: Arend van Spriel <aspriel.gmail.com>
Reviewed-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://lore.kernel.org/r/id1HN6qCMAirApBzTA6fT7ZFWBBGCJhULpflxQ7NT6cgCboVnn3RHpiOFjA9SbRqzBRFLk9ES0C4FNvO6fUQsNg7pqF6ZSNAYUo99nHy8PY=@dannyvanheumen.nl
drivers/net/wireless/broadcom/brcm80211/brcmfmac/bcmsdh.c
drivers/net/wireless/broadcom/brcm80211/brcmfmac/sdio.c