Re: [BUG] Unloading mt7921e module cause use-after-free

[Mikhail Gavrilov]

On Fri, Jan 12, 2024 at 1:06 PM Mikhail Gavrilov
<mikhail.v.gavrilov@xxxxxxxxx> wrote:
>
> On Fri, Jan 12, 2024 at 12:31 PM Mikhail Gavrilov
> <mikhail.v.gavrilov@xxxxxxxxx> wrote:
> >
> > Thanks, this patch looks good to me.
> > Demonstration: https://youtu.be/nKnA2ftVoXw
> >
> > Tested-by: Mikhail Gavrilov <mikhail.v.gavrilov@xxxxxxxxx>
> >
>
> I noticed DMA-API notice:
> ------------[ cut here ]------------
> DMA-API: pci 0000:0f:00.0: device driver has pending DMA allocations
> while released from device [count=21]
> One of leaked entries details: [device address=0x00000000ffbda000]
> [size=4096 bytes] [mapped with DMA_FROM_DEVICE] [mapped as single]
> WARNING: CPU: 13 PID: 11252 at kernel/dma/debug.c:863
> dma_debug_device_change+0x276/0x3d0
> Modules linked in: mt7921e(-) uinput rfcomm snd_seq_dummy snd_hrtimer
> nf_conntrack_netbios_ns nf_conntrack_broadcast nft_fib_inet
> nft_fib_ipv4 nft_fib_ipv6 nft_fib nft_reject_inet nf_reject_ipv4
> nft_reject nft_ct nft_chain_nat nf_nat nf_conntrack nf_defrag_ipv6
> nf_defrag_ipv4 ip_set nf_tables qrtr bnep sunrpc binfmt_misc
> intel_rapl_msr intel_rapl_common mt7921_common mt792x_lib
> snd_hda_codec_hdmi mt76 edac_mce_amd snd_hda_intel snd_intel_dspcfg
> snd_intel_sdw_acpi mac80211 snd_usb_audio kvm_amd snd_hda_codec
> snd_usbmidi_lib btusb snd_hda_core uvcvideo snd_ump btrtl kvm ntel
> snd_hwdep btbcm uvc snd_seq vfat btmtk videobuf2_vmalloc
> videobuf2_memops libarc4 fat snd_seq_device videobuf2_v4l2 bluetooth
> snd_pcm irqbypass videobuf2_common snd_timer cfg80211 rapl of snd
> joydev pcspkr mc k10temp i2c_piix4 soundcore rfkill gpio_amdpt
> gpio_generic loop nfnetlink zram amdgpu amdxcp i2c_algo_bit
> drm_ttm_helper ttm drm_exec crct10dif_pclmul gpu_sched
>  crc32_pclmul crc32c_intel drm_suballoc_helper polyval_clmulni
> drm_buddy polyval_generic drm_display_helper ghash_clmulni_intel nvme
> sha512_ssse3 sha256_ssse3 ccp nvme_core sha1_ssse3 sp5100_tco ek
> nvme_auth video wmi ip6_tables ip_tables fuse [last unloaded: mt7921e]
> CPU: 13 PID: 11252 Comm: rmmod Tainted: G        W    L     6.7.0-check-fix+ #70
> Hardware name: Micro-Star International Co., Ltd. MS-7D73/MPG B650I
> EDGE WIFI (MS-7D73), BIOS 1.81 01/05/2024
> RIP: 0010:dma_debug_device_change+0x276/0x3d0
> Code: 54 24 08 e8 5c d7 c8 01 48 8b 54 24 08 48 89 c6 ff 34 24 49 89
> e9 49 89 d8 44 89 e1 41 56 48 c7 c7 40 46 52 b3 e8 1a c9 d6 ff <0f> 0b
> 5a 59 4d 85 ed 74 49 48 c7 c7 80 47 52 b3 e8 15 04 f5 ff
> RSP: 0018:ffffc90009c77bf0 EFLAGS: 00010286
> RAX: 0000000000000000 RBX: 00000000ffbda000 RCX: 0000000000000000
> RDX: 0000000000000002 RSI: 0000000000000004 RDI: 0000000000000001
> RBP: 0000000000001000 R08: 0000000000000001 R09: ffffed11f587fff9
> R10: ffff888fac3fffcb R11: 0000000000000000 R12: 0000000000000015
> R13: ffff888109b02180 R14: ffffffffb35265e0 R15: ffff88811418e0c0
> FS:  00007f5a1e4177c0(0000) GS:ffff888fac200000(0000) knlGS:0000000000000000
> CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> CR2: 00007f8b1d394054 CR3: 0000000134cb8000 CR4: 0000000000f50ef0
> PKRU: 55555554
> Call Trace:
>  <TASK>
>  ? __warn+0xcd/0x2b0
>  ? __pfx_vprintk_emit+0x10/0x10
>  ? dma_debug_device_change+0x276/0x3d0
>  ? report_bug+0x2ea/0x390
>  ? handle_bug+0x79/0xa0
>  ? exc_invalid_op+0x17/0x40
>  ? asm_exc_invalid_op+0x1a/0x20
>  ? dma_debug_device_change+0x276/0x3d0
>  notifier_call_chain+0xa0/0x2a0
>  blocking_notifier_call_chain+0x64/0x90
>  bus_notify+0x51/0x70
>  device_release_driver_internal+0x42d/0x540
>  driver_detach+0xc5/0x180
>  bus_remove_driver+0x11e/0x2a0
>  ? __check_object_size+0x5b/0x680
>  pci_unregister_driver+0x2a/0x250
>  __do_sys_delete_module+0x350/0x580
>  ? __pfx___do_sys_delete_module+0x10/0x10
>  ? syscall_exit_to_user_mode+0xce/0x2b0
>  ? do_syscall_64+0xab/0x190
>  ? rcu_is_watching+0x15/0xb0
>  ? syscall_exit_to_user_mode+0xb6/0x2b0
>  ? trace_hardirqs_on_prepare+0xe3/0x100
>  ? do_syscall_64+0x58/0x190
>  do_syscall_64+0x9b/0x190
>  ? trace_hardirqs_on_prepare+0xe3/0x100
>  entry_SYSCALL_64_after_hwframe+0x6e/0x76
> RIP: 0033:0x7f5a1dd2d05b
> Code: 73 01 c3 48 8b 0d bd 8d 0c 00 f7 d8 64 89 01 48 83 c8 ff c3 66
> 2e 0f 1f 84 00 00 00 00 00 90 f3 0f 1e fa b8 b0 00 00 00 0f 05 <48> 3d
> 01 f0 ff ff 73 01 c3 48 8b 0d 8d 8d 0c 00 f7 d8 64 89 01
> RSP: 002b:00007ffc5ae71b68 EFLAGS: 00000206 ORIG_RAX: 00000000000000b0
> RAX: ffffffffffffffda RBX: 000055e490fee760 RCX: 00007f5a1dd2d05b
> RDX: 0000000000000000 RSI: 0000000000000800 RDI: 000055e490fee7c8
> RBP: 00007ffc5ae71b90 R08: 1999999999999999 R09: 0000000000000000
> R10: 00007f5a1dda5ac0 R11: 0000000000000206 R12: 0000000000000000
> R13: 00007ffc5ae71df0 R14: 000055e490fee760 R15: 0000000000000000
>  </TASK>
> irq event stamp: 0
> hardirqs last  enabled at (0): [<0000000000000000>] 0x0
> hardirqs last disabled at (0): [<ffffffffb023b291>] copy_process+0x2111/0x88e0
> softirqs last  enabled at (0): [<ffffffffb023b2f3>] copy_process+0x2173/0x88e0
> softirqs last disabled at (0): [<0000000000000000>] 0x0
> ---[ end trace 0000000000000000 ]---
> DMA-API: Mapped at:
>  debug_dma_map_page+0x60/0x3c0
>  dma_map_page_attrs+0x2fc/0xba0
>  page_pool_dma_map+0xaf/0x2d0
>  __page_pool_alloc_pages_slow+0x36f/0xab0
>  page_pool_alloc_frag+0x4fa/0x9c0
>
> It happens when I unload driver mt7921e and in the background some
> applications (in my case it is speedtest) have heavy network activity.
>
> Demonstration: https://youtu.be/XYO4ueVlh90
>
> But I am not sure what it is related to the mt7921e driver.
>

Deren, excuse me.
Do you plan to send this patch to the stable branch?
Or will you find another approach?

-- 
Best Regards,
Mike Gavrilov.