Re: [PATCH] mmc: queue:Improve error handling during allocation of bounce buffers.
From: Ulf Hansson
Date: Wed Dec 03 2014 - 07:36:45 EST
On 3 December 2014 at 11:36, Bhuvanesh <bhuvanesh_surachari@xxxxxxxxxx> wrote:
> Hi,
>
> During our rigorous testing of inserting and removing SD card we found
> exception in the kernel. Please find the backtrace as below:
>
> [ 1605.392278] Backtrace:
> [ 1605.395466] [<800117c4>] (dump_backtrace+0x0/0x100) from [<803f2cf4>] (dump_stack+0x18/0x1c)
> [ 1605.408679] [<803f2cdc>] (dump_stack+0x0/0x1c) from [<800b4038>] (warn_alloc_failed+0xec/0x10c)
> [ 1605.423194] [<800b3f4c>] (warn_alloc_failed+0x0/0x10c) from [<800b6d10>] (__alloc_pages_nodemask+0x764/0x890)
> [ 1605.439126] [<800b65ac>] (__alloc_pages_nodemask+0x0/0x890) from [<800b6e54>] (__get_free_pages+0x18/0x54)
> [ 1605.453885] [<800b6e3c>] (__get_free_pages+0x0/0x54) from [<800e4504>] (kmalloc_order_trace+0x2c/0xe8)
> [ 1605.470367] [<800e44d8>] (kmalloc_order_trace+0x0/0xe8) from [<800e5c4c>] (__kmalloc+0x38/0x1e4)
> [ 1605.482358] [<800e5c14>] (__kmalloc+0x0/0x1e4) from [<8026ac50>] (mmc_init_queue+0x198/0x444)
> [ 1605.494663] [<8026aab8>] (mmc_init_queue+0x0/0x444) from [<80268638>] (mmc_blk_alloc_req+0x184/0x354)
> [ 1605.513224] [<802684b4>] (mmc_blk_alloc_req+0x0/0x354) from [<80268f54>] (mmc_blk_probe+0x7c/0x28c)
> [ 1605.527459] [<80268ed8>] (mmc_blk_probe+0x0/0x28c) from [<8025e8f0>] (mmc_bus_probe+0x1c/0x20)
> [ 1605.543524] [<8025e8d4>] (mmc_bus_probe+0x0/0x20) from [<802a0d2c>] (driver_probe_device+0xb4/0x204)
> [ 1605.558819] [<802a0c78>] (driver_probe_device+0x0/0x204) from [<802a0eac>] (__device_attach+0x30/0x4c)
> [ 1605.571747] [<802a0e7c>] (__device_attach+0x0/0x4c) from [<8029f3b4>] (bus_for_each_drv+0x80/0x94)
> [ 1605.587027] [<8029f334>] (bus_for_each_drv+0x0/0x94) from [<802a0c2c>] (device_attach+0x70/0x94)
> [ 1605.601680] [<802a0bbc>] (device_attach+0x0/0x94) from [<802a01e0>] (bus_probe_device+0x30/0xa0)
> [ 1605.614128] [<802a01b0>] (bus_probe_device+0x0/0xa0) from [<8029e8bc>] (device_add+0x42c/0x570)
> [ 1605.626933] [<8029e490>] (device_add+0x0/0x570) from [<8025eda0>] (mmc_add_card+0x188/0x1e4)
> [ 1605.638880] [<8025ec18>] (mmc_add_card+0x0/0x1e4) from [<802637bc>] (mmc_attach_sd+0x188/0x210)
> [ 1605.651413] [<80263634>] (mmc_attach_sd+0x0/0x210) from [<8025e508>] (mmc_rescan+0x240/0x2ac)
> [ 1605.666968] [<8025e2c8>] (mmc_rescan+0x0/0x2ac) from [<8003cde0>] (process_one_work+0x2cc/0x460)
> [ 1605.682229] [<8003cb14>] (process_one_work+0x0/0x460) from [<8003d240>] (worker_thread+0x298/0x3ec)
> [ 1605.698223] [<8003cfa8>] (worker_thread+0x0/0x3ec) from [<80042490>] (kthread+0xb4/0xc0)
> [ 1605.707312] [<800423dc>] (kthread+0x0/0xc0) from [<8000d9b8>] (ret_from_fork+0x14/0x3c)
> [ 1605.846505] Unable to handle kernel NULL pointer dereference at virtual address 00000000
> [ 1605.859027] pgd = 80004000
> [ 1605.862280] [00000000] *pgd=00000000
> [ 1605.866415] Internal error: Oops: 17 [#1] PREEMPT SMP ARM
> [ 1605.949143] CPU: 0 Not tainted (3.8.13.27-03391-g4e6f494 #1)
> [ 1605.955216] PC is at __blk_segment_map_sg+0xfc/0x140
> [ 1605.955216] PC is at __blk_segment_map_sg+0xfc/0x140
> [ 1605.960240] LR is at blk_rq_map_sg+0xa0/0x1d8
> [ 1605.964657] pc : [<802240fc>] lr : [<802241e0>] psr: 60000013
> [ 1605.964657] sp : 81fe5db8 ip : 81fe5df0 fp : 81fe5dec
> [ 1605.976191] r10: 81fe5e04 r9 : 00000000 r8 : 00000000
> [ 1605.981473] r7 : 81fe5e00 r6 : 00001000 r5 : 81fe5e04 r4 : ad1ea3c4
> [ 1605.988056] r3 : 00000000 r2 : 80667000 r1 : 00000000 r0 : ac2972a0
> [ 1605.994638] Flags: nZCv IRQs on FIQs on Mode SVC_32 ISA ARM Segment kernel
> [ 1606.002001] Control: 10c5387d Table: 3d09004a DAC: 00000015
> [ 1606.007801] Process mmcqd/0 (pid: 13804, stack limit = 0x81fe4238)
> [ 1606.014034] Stack: (0x81fe5db8 to 0x81fe6000)
> [ 1606.018446] 5da0: ac274a38 00000000
> [ 1606.026681] 5dc0: 00000000 ac2972a0 ad1f4af8 ad1ea380 00000001 00000000 00000000 81fe5e04
> [ 1606.034917] 5de0: 81fe5e3c 81fe5df0 802241e0 8022400c 81fe5e04 81fe5e08 81fe5e0c 00000031
> [ 1606.043152] 5e00: 00000000 00000000 00000000 00000001 81fe5e54 8633a030 8633a800 ad1f4af8
> [ 1606.051386] 5e20: 00000008 8633a034 8633a008 8633a000 81fe5e5c 81fe5e40 8026b06c 8022414c
> [ 1606.059622] 5e40: 8633a030 8633a800 ad1f4af8 00000008 81fe5e94 81fe5e60 80268c14 8026b04c
> [ 1606.067857] 5e60: 803f90d8 00000000 81fe5e8c 8633a800 ad1f4af8 8633a008 ac07f000 ad1f4af8
> [ 1606.076091] 5e80: 8633a000 8633a034 81fe5edc 81fe5e98 802699a8 80268994 81fe5eb4 00000000
> [ 1606.084326] 5ea0: 00000000 00000000 00100100 00200200 803f7f98 ad1f4af8 8633a800 8633a008
> [ 1606.092562] 5ec0: ac07f000 8633a000 81fe4000 8633a000 81fe5f24 81fe5ee0 8026a210 80269920
> [ 1606.100798] 5ee0: 00000000 ad1f4af8 81fe5f0c 81fe5ef8 802204e0 120d0000 120d0000 8633a008
> [ 1606.109034] 5f00: ad1f4af8 8633a010 ac2972a0 81fe4028 81fe4000 00000001 81fe5f5c 81fe5f28
> [ 1606.117270] 5f20: 8026a89c 80269d4c 00000000 120d0000 8026a7bc ac23bc28 00000000 8633a008
> [ 1606.125506] 5f40: 8026a7bc 00000000 00000000 00000000 81fe5fac 81fe5f60 80042490 8026a7c8
> [ 1606.133742] 5f60: 803f9134 00000000 81fe5f94 8633a008 00000000 00000000 81fe5f78 81fe5f78
> [ 1606.141977] 5f80: 00000000 00000000 81fe5f88 81fe5f88 ac23bc28 800423dc 00000000 00000000
> [ 1606.150213] 5fa0: 00000000 81fe5fb0 8000d9b8 800423e8 00000000 00000000 00000000 00000000
> [ 1606.158448] 5fc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
> [ 1606.166684] 5fe0: 00000000 00000000 00000000 00000000 00000013 00000000 1da1d629 0030d97d
> [ 1606.174914] Backtrace:
> [ 1606.177445] [<80224000>] (__blk_segment_map_sg+0x0/0x140) from [<802241e0>] (blk_rq_map_sg+0xa0/0x1d8)
> [ 1606.186822] [<80224140>] (blk_rq_map_sg+0x0/0x1d8) from [<8026b06c>] (mmc_queue_map_sg+0x2c/0x94)
> [ 1606.195759] [<8026b040>] (mmc_queue_map_sg+0x0/0x94) from [<80268c14>] (mmc_blk_rw_rq_prep+0x28c/0x300)
> [ 1606.210939] [<80268988>] (mmc_blk_rw_rq_prep+0x0/0x300) from [<802699a8>] (mmc_blk_issue_rw_rq+0x94/0x42c)
> [ 1606.220655] [<80269914>] (mmc_blk_issue_rw_rq+0x0/0x42c) from [<8026a210>] (mmc_blk_issue_rq+0x4d0/0x500)
> [ 1606.230284] [<80269d40>] (mmc_blk_issue_rq+0x0/0x500) from [<8026a89c>] (mmc_queue_thread+0xe0/0x17c)
> [ 1606.239571] [<8026a7bc>] (mmc_queue_thread+0x0/0x17c) from [<80042490>] (kthread+0xb4/0xc0)
> [ 1606.247989] [<800423dc>] (kthread+0x0/0xc0) from [<8000d9b8>] (ret_from_fork+0x14/0x3c)
> [ 1606.261771] Code: e5850000 e5953000 e5942000 e5941008 (e5930000)
> [ 1606.301708] Kernel panic - not syncing: Fatal exception
> [ 1606.307005] Rebooting in 1 seconds..
>
> On analysis i found that error handling in mmc_init_queue during
> allocation of bounce buffers is incomplete. In the exception case the
> allocation of current bounce buffer fails and the previous bounce
> buffer allocation succeeds. This later leads to exception in,
> __blk_segment_map_sg while trying to access the bounce buffer.
>
> The below patch improves the error handling during allocation of
> bounce buffers. The previous bounce buffer is allocated only if the
> allocation of current bounce buffer succeeds.
>
>
> From 7ea020e32e1fbe4d05d104e31815d908af92f2a5 Mon Sep 17 00:00:00 2001
> From: Bhuvanesh Surachari <bhuvanesh_surachari@xxxxxxxxxx>
> Date: Mon, 1 Dec 2014 02:23:02 -0500
> Subject: [PATCH] mmc: queue:Improve error handling during allocation of
> bounce buffers.
>
> Allocation of previous bounce buffer in mmc_init_queue when the current
> bounce buffer allocation fails was leading to a crash later in
> __blk_segment_map_sg. Error handling is improved by allocating previous
> bounce buffer only if the current bounce buffer allocation succeeds.
>
> Signed-off-by: Bhuvanesh Surachari <bhuvanesh_surachari@xxxxxxxxxx>
> Signed-off-by: Harish Jenny K N <harish_kandiga@xxxxxxxxxx>
Nope, this is still not the correct way to send a patch. Everything
above will be treated as the commit message.
> ---
If you want to send a message as a part of your patch add it in the
patch file, after these three dashes.
End the message, by adding another three dashes on a separate line.
> drivers/mmc/card/queue.c | 16 +++++++++-------
> 1 file changed, 9 insertions(+), 7 deletions(-)
>
> diff --git a/drivers/mmc/card/queue.c b/drivers/mmc/card/queue.c
> index cfa6110..236d194 100644
> --- a/drivers/mmc/card/queue.c
> +++ b/drivers/mmc/card/queue.c
> @@ -232,13 +232,15 @@ int mmc_init_queue(struct mmc_queue *mq, struct mmc_card *card,
> if (!mqrq_cur->bounce_buf) {
> pr_warn("%s: unable to allocate bounce cur buffer\n",
> mmc_card_name(card));
> - }
> - mqrq_prev->bounce_buf = kmalloc(bouncesz, GFP_KERNEL);
> - if (!mqrq_prev->bounce_buf) {
> - pr_warn("%s: unable to allocate bounce prev buffer\n",
> - mmc_card_name(card));
> - kfree(mqrq_cur->bounce_buf);
> - mqrq_cur->bounce_buf = NULL;
> + } else {
> + mqrq_prev->bounce_buf =
> + kmalloc(bouncesz, GFP_KERNEL);
> + if (!mqrq_prev->bounce_buf) {
> + pr_warn("%s: unable to allocate bounce prev buffer\n",
> + mmc_card_name(card));
> + kfree(mqrq_cur->bounce_buf);
> + mqrq_cur->bounce_buf = NULL;
> + }
> }
> }
>
> --
> 1.7.9.5
>
>
> Kindly review the patch.
>
> Thank you,
> Regards,
> Bhuvanesh
Kind regards
Uffe
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/