342-v5.2-0001-brcmfmac-fix-race-during-disconnect-when-USB-complet.patch 4.9 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384
  1. From db3b9e2e1d58080d0754bdf9293dabf8c6491b67 Mon Sep 17 00:00:00 2001
  2. From: Piotr Figiel <p.figiel@camlintechnologies.com>
  3. Date: Fri, 8 Mar 2019 15:25:04 +0000
  4. Subject: [PATCH] brcmfmac: fix race during disconnect when USB completion is
  5. in progress
  6. It was observed that rarely during USB disconnect happening shortly after
  7. connect (before full initialization completes) usb_hub_wq would wait
  8. forever for the dev_init_lock to be unlocked. dev_init_lock would remain
  9. locked though because of infinite wait during usb_kill_urb:
  10. [ 2730.656472] kworker/0:2 D 0 260 2 0x00000000
  11. [ 2730.660700] Workqueue: events request_firmware_work_func
  12. [ 2730.664807] [<809dca20>] (__schedule) from [<809dd164>] (schedule+0x4c/0xac)
  13. [ 2730.670587] [<809dd164>] (schedule) from [<8069af44>] (usb_kill_urb+0xdc/0x114)
  14. [ 2730.676815] [<8069af44>] (usb_kill_urb) from [<7f258b50>] (brcmf_usb_free_q+0x34/0xa8 [brcmfmac])
  15. [ 2730.684833] [<7f258b50>] (brcmf_usb_free_q [brcmfmac]) from [<7f2517d4>] (brcmf_detach+0xa0/0xb8 [brcmfmac])
  16. [ 2730.693557] [<7f2517d4>] (brcmf_detach [brcmfmac]) from [<7f251a34>] (brcmf_attach+0xac/0x3d8 [brcmfmac])
  17. [ 2730.702094] [<7f251a34>] (brcmf_attach [brcmfmac]) from [<7f2587ac>] (brcmf_usb_probe_phase2+0x468/0x4a0 [brcmfmac])
  18. [ 2730.711601] [<7f2587ac>] (brcmf_usb_probe_phase2 [brcmfmac]) from [<7f252888>] (brcmf_fw_request_done+0x194/0x220 [brcmfmac])
  19. [ 2730.721795] [<7f252888>] (brcmf_fw_request_done [brcmfmac]) from [<805748e4>] (request_firmware_work_func+0x4c/0x88)
  20. [ 2730.731125] [<805748e4>] (request_firmware_work_func) from [<80141474>] (process_one_work+0x228/0x808)
  21. [ 2730.739223] [<80141474>] (process_one_work) from [<80141a80>] (worker_thread+0x2c/0x564)
  22. [ 2730.746105] [<80141a80>] (worker_thread) from [<80147bcc>] (kthread+0x13c/0x16c)
  23. [ 2730.752227] [<80147bcc>] (kthread) from [<801010b4>] (ret_from_fork+0x14/0x20)
  24. [ 2733.099695] kworker/0:3 D 0 1065 2 0x00000000
  25. [ 2733.103926] Workqueue: usb_hub_wq hub_event
  26. [ 2733.106914] [<809dca20>] (__schedule) from [<809dd164>] (schedule+0x4c/0xac)
  27. [ 2733.112693] [<809dd164>] (schedule) from [<809e2a8c>] (schedule_timeout+0x214/0x3e4)
  28. [ 2733.119621] [<809e2a8c>] (schedule_timeout) from [<809dde2c>] (wait_for_common+0xc4/0x1c0)
  29. [ 2733.126810] [<809dde2c>] (wait_for_common) from [<7f258d00>] (brcmf_usb_disconnect+0x1c/0x4c [brcmfmac])
  30. [ 2733.135206] [<7f258d00>] (brcmf_usb_disconnect [brcmfmac]) from [<8069e0c8>] (usb_unbind_interface+0x5c/0x1e4)
  31. [ 2733.143943] [<8069e0c8>] (usb_unbind_interface) from [<8056d3e8>] (device_release_driver_internal+0x164/0x1fc)
  32. [ 2733.152769] [<8056d3e8>] (device_release_driver_internal) from [<8056c078>] (bus_remove_device+0xd0/0xfc)
  33. [ 2733.161138] [<8056c078>] (bus_remove_device) from [<8056977c>] (device_del+0x11c/0x310)
  34. [ 2733.167939] [<8056977c>] (device_del) from [<8069cba8>] (usb_disable_device+0xa0/0x1cc)
  35. [ 2733.174743] [<8069cba8>] (usb_disable_device) from [<8069507c>] (usb_disconnect+0x74/0x1dc)
  36. [ 2733.181823] [<8069507c>] (usb_disconnect) from [<80695e88>] (hub_event+0x478/0xf88)
  37. [ 2733.188278] [<80695e88>] (hub_event) from [<80141474>] (process_one_work+0x228/0x808)
  38. [ 2733.194905] [<80141474>] (process_one_work) from [<80141a80>] (worker_thread+0x2c/0x564)
  39. [ 2733.201724] [<80141a80>] (worker_thread) from [<80147bcc>] (kthread+0x13c/0x16c)
  40. [ 2733.207913] [<80147bcc>] (kthread) from [<801010b4>] (ret_from_fork+0x14/0x20)
  41. It was traced down to a case where usb_kill_urb would be called on an URB
  42. structure containing more or less random data, including large number in
  43. its use_count. During the debugging it appeared that in brcmf_usb_free_q()
  44. the traversal over URBs' lists is not synchronized with operations on those
  45. lists in brcmf_usb_rx_complete() leading to handling
  46. brcmf_usbdev_info structure (holding lists' head) as lists' element and in
  47. result causing above problem.
  48. Fix it by walking through all URBs during brcmf_cancel_all_urbs using the
  49. arrays of requests instead of linked lists.
  50. Signed-off-by: Piotr Figiel <p.figiel@camlintechnologies.com>
  51. Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
  52. ---
  53. drivers/net/wireless/broadcom/brcm80211/brcmfmac/usb.c | 10 ++++++++--
  54. 1 file changed, 8 insertions(+), 2 deletions(-)
  55. --- a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/usb.c
  56. +++ b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/usb.c
  57. @@ -684,12 +684,18 @@ static int brcmf_usb_up(struct device *d
  58. static void brcmf_cancel_all_urbs(struct brcmf_usbdev_info *devinfo)
  59. {
  60. + int i;
  61. +
  62. if (devinfo->ctl_urb)
  63. usb_kill_urb(devinfo->ctl_urb);
  64. if (devinfo->bulk_urb)
  65. usb_kill_urb(devinfo->bulk_urb);
  66. - brcmf_usb_free_q(&devinfo->tx_postq, true);
  67. - brcmf_usb_free_q(&devinfo->rx_postq, true);
  68. + if (devinfo->tx_reqs)
  69. + for (i = 0; i < devinfo->bus_pub.ntxq; i++)
  70. + usb_kill_urb(devinfo->tx_reqs[i].urb);
  71. + if (devinfo->rx_reqs)
  72. + for (i = 0; i < devinfo->bus_pub.nrxq; i++)
  73. + usb_kill_urb(devinfo->rx_reqs[i].urb);
  74. }
  75. static void brcmf_usb_down(struct device *dev)