| Message ID | 1a5810475d7a475db5e4e5130b8f455c@hyperstone.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [PATCHv2] mmc: core: fix race of queue reset and card removal | expand |
I guess treat both versions more like RFCs. V1 seems a bit nicer, v2 avoids the dangling mq->card, which I don't see an issue with but am easily convinced otherwise. Both fix the issue for me. -----Original Message----- From: Christian Löhle <CLoehle@hyperstone.com> Sent: Dienstag, 4. Oktober 2022 15:14 To: ulf.hansson@linaro.org; Adrian Hunter <adrian.hunter@intel.com>; Linux MMC List <linux-mmc@vger.kernel.org>; linux-kernel@vger.kernel.org; Christian Löhle <CLoehle@hyperstone.com> Cc: Avri Altman <Avri.Altman@wdc.com> Subject: [PATCHv2] mmc: core: fix race of queue reset and card removal If a recovery is active and the card is removed do not try to switch back partitions. Furthermore do not reference mq->card which might be NULLed in the meantime. This has been observed with recovery active with CQE. [ 1083.510578] Unable to handle kernel NULL pointer dereference at virtual address 000000000000038c [ 1083.511362] Mem abort info: [ 1083.511626] ESR = 0x96000004 [ 1083.511912] EC = 0x25: DABT (current EL), IL = 32 bits [ 1083.512395] SET = 0, FnV = 0 [ 1083.512681] EA = 0, S1PTW = 0 [ 1083.512973] FSC = 0x04: level 0 translation fault [ 1083.513417] Data abort info: [ 1083.513686] ISV = 0, ISS = 0x00000004 [ 1083.514039] CM = 0, WnR = 0 [ 1083.514318] user pgtable: 4k pages, 48-bit VAs, pgdp=000000000a4c3000 [ 1083.514899] [000000000000038c] pgd=0000000000000000, p4d=0000000000000000 [ 1083.515854] Internal error: Oops: 96000004 [#1] SMP [ 1083.516295] CPU: 0 PID: 153 Comm: kworker/0:2 Tainted: G W 5.18.12-g925ff1d10c99-dirty #7 [ 1083.517127] Hardware name: Pine64 RockPro64 v2.1 (DT) [ 1083.517574] Workqueue: events mmc_mq_recovery_handler [ 1083.518032] pstate: 60000005 (nZCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--) [ 1083.518645] pc : mmc_blk_reset+0x60/0x1ac [ 1083.519004] lr : mmc_blk_reset+0x38/0x1ac [ 1083.519361] sp : ffff8000100b3cd0 [ 1083.519654] x29: ffff8000100b3cd0 x28: 0000000000000000 x27: 0000000000000000 [ 1083.520288] x26: ffff80000b0ba000 x25: ffff0000f6e74805 x24: ffff000004c2fdc0 [ 1083.520922] x23: ffff000014950000 x22: ffff000004c2fc18 x21: ffff00000a33c000 [ 1083.521556] x20: 00000000ffffff85 x19: ffff000004c2fc00 x18: ffffffffffffffff [ 1083.522189] x17: ffff80000cd9b200 x16: ffff80000cd9b190 x15: 0000000000000006 [ 1083.522823] x14: 0000000000000000 x13: ffff80000b0c28f0 x12: 0000000000001707 [ 1083.523457] x11: 00000000000007ad x10: ffff80000c6c28f0 x9 : ffff80000b0c28f0 [ 1083.524090] x8 : 00000000fffbffff x7 : 0000000000000001 x6 : 0000000000000000 [ 1083.524723] x5 : 0000000000000000 x4 : ffff0000f6e62d30 x3 : 0000000000000000 [ 1083.525357] x2 : 0000000000000000 x1 : ffff00000b6e0000 x0 : 0000000000000000 [ 1083.525990] Call trace: [ 1083.526209] mmc_blk_reset+0x60/0x1ac [ 1083.526536] mmc_blk_cqe_recovery+0x8c/0xd0 [ 1083.526908] mmc_mq_recovery_handler+0xc4/0xd0 [ 1083.527303] process_one_work+0x23c/0x3fc [ 1083.527663] worker_thread+0x74/0x420 [ 1083.527990] kthread+0xec/0xf0 [ 1083.528264] ret_from_fork+0x10/0x20 [ 1083.528587] Code: d50323bf d65f03c0 f94352a0 f9404000 (b9438c01) [ 1083.529126] ---[ end trace 0000000000000000 ]--- [ 1431.677970] Unable to handle kernel NULL pointer dereference at virtual address 0000000000000000 [ 1431.678753] Mem abort info: [ 1431.679017] ESR = 0x96000004 [ 1431.679303] EC = 0x25: DABT (current EL), IL = 32 bits [ 1431.679786] SET = 0, FnV = 0 [ 1431.680072] EA = 0, S1PTW = 0 [ 1431.680366] FSC = 0x04: level 0 translation fault [ 1431.680810] Data abort info: [ 1431.681080] ISV = 0, ISS = 0x00000004 [ 1431.681432] CM = 0, WnR = 0 [ 1431.681712] user pgtable: 4k pages, 48-bit VAs, pgdp=000000000bb98000 [ 1431.682390] [0000000000000000] pgd=0000000000000000, p4d=0000000000000000 [ 1431.683393] Internal error: Oops: 96000004 [#1] SMP [ 1431.683841] CPU: 0 PID: 19948 Comm: kworker/0:2 Not tainted 5.18.12-gf65532578f32-dirty #16 [ 1431.684576] Hardware name: Pine64 RockPro64 v2.1 (DT) [ 1431.685024] Workqueue: events mmc_mq_recovery_handler [ 1431.685487] pstate: 60000005 (nZCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--) [ 1431.686100] pc : mmc_put_card+0x38/0x110 [ 1431.686453] lr : mmc_mq_recovery_handler+0x98/0xd0 [ 1431.686879] sp : ffff800015813cf0 [ 1431.687173] x29: ffff800015813cf0 x28: 0000000000000000 x27: 0000000000000000 [ 1431.687807] x26: ffff80000b0ba000 x25: ffff0000f6e74805 x24: ffff000013bd65c0 [ 1431.688441] x23: ffff000013b96120 x22: ffff000013bd6418 x21: 0000000000000000 [ 1431.689075] x20: ffff800008ed1c70 x19: ffff8000091767d8 x18: ffffffffffffffff [ 1431.689709] x17: 31335b1b6d375b1b x16: 6d305b1b47554245 x15: 0000000000000006 [ 1431.690343] x14: 0000000000000000 x13: 0000000000000000 x12: 0000000000000000 [ 1431.690976] x11: ffff000013bd6570 x10: 0000000000000001 x9 : ffff80000ea69228 [ 1431.691611] x8 : ffff80000df892c8 x7 : 0000000000000000 x6 : 0000000000000001 [ 1431.692245] x5 : 0000000000000001 x4 : 0000000000000002 x3 : ffff80000e6feac8 [ 1431.692879] x2 : 000000000000036e x1 : ffff800008ed1c70 x0 : 0000000000000000 [ 1431.693513] Call trace: [ 1431.693732] mmc_put_card+0x38/0x110 [ 1431.694055] mmc_mq_recovery_handler+0x98/0xd0 [ 1431.694452] process_one_work+0x23c/0x3fc [ 1431.694812] worker_thread+0x74/0x420 [ 1431.695139] kthread+0xec/0xf0 [ 1431.695414] ret_from_fork+0x10/0x20 [ 1431.695738] Code: f9001bf7 aa0103f6 aa0003f5 aa1403e1 (f9400017) [ 1431.696278] ---[ end trace 0000000000000000 ]--- Signed-off-by: Christian Loehle <cloehle@hyperstone.com> --- drivers/mmc/core/block.c | 4 ++-- drivers/mmc/core/queue.c | 5 +++-- 2 files changed, 5 insertions(+), 4 deletions(-) diff --git a/drivers/mmc/core/block.c b/drivers/mmc/core/block.c index ce89611a136e..0cd3a7065629 100644 --- a/drivers/mmc/core/block.c +++ b/drivers/mmc/core/block.c @@ -997,8 +997,8 @@ static int mmc_blk_reset(struct mmc_blk_data *md, struct mmc_host *host, md->reset_done |= type; err = mmc_hw_reset(host->card); - /* Ensure we switch back to the correct partition */ - if (err) { + /* Ensure we switch back to the correct partition on successful reset */ + if (!err) { struct mmc_blk_data *main_md = dev_get_drvdata(&host->card->dev); int part_err; diff --git a/drivers/mmc/core/queue.c b/drivers/mmc/core/queue.c index fefaa901b50f..6931fa082ea7 100644 --- a/drivers/mmc/core/queue.c +++ b/drivers/mmc/core/queue.c @@ -137,9 +137,10 @@ static void mmc_mq_recovery_handler(struct work_struct *work) struct mmc_queue *mq = container_of(work, struct mmc_queue, recovery_work); struct request_queue *q = mq->queue; + struct mmc_card *card = mq->card; struct mmc_host *host = mq->card->host; - mmc_get_card(mq->card, &mq->ctx); + mmc_get_card(card, &mq->ctx); mq->in_recovery = true; @@ -157,7 +158,7 @@ static void mmc_mq_recovery_handler(struct work_struct *work) if (host->hsq_enabled) host->cqe_ops->cqe_recovery_finish(host); - mmc_put_card(mq->card, &mq->ctx); + mmc_put_card(card, &mq->ctx); blk_mq_run_hw_queues(q, true); } -- 2.37.3 Hyperstone GmbH | Reichenaustr. 39a | 78467 Konstanz Managing Director: Dr. Jan Peter Berns. Commercial register of local courts: Freiburg HRB381782
On 4/10/22 16:13, Christian Löhle wrote: > If a recovery is active and the card is removed do not > try to switch back partitions. Furthermore do not reference > mq->card which might be NULLed in the meantime. > > This has been observed with recovery active with CQE. > [ 1083.510578] Unable to handle kernel NULL pointer dereference at virtual address 000000000000038c > [ 1083.511362] Mem abort info: > [ 1083.511626] ESR = 0x96000004 > [ 1083.511912] EC = 0x25: DABT (current EL), IL = 32 bits > [ 1083.512395] SET = 0, FnV = 0 > [ 1083.512681] EA = 0, S1PTW = 0 > [ 1083.512973] FSC = 0x04: level 0 translation fault > [ 1083.513417] Data abort info: > [ 1083.513686] ISV = 0, ISS = 0x00000004 > [ 1083.514039] CM = 0, WnR = 0 > [ 1083.514318] user pgtable: 4k pages, 48-bit VAs, pgdp=000000000a4c3000 > [ 1083.514899] [000000000000038c] pgd=0000000000000000, p4d=0000000000000000 > [ 1083.515854] Internal error: Oops: 96000004 [#1] SMP > [ 1083.516295] CPU: 0 PID: 153 Comm: kworker/0:2 Tainted: G W 5.18.12-g925ff1d10c99-dirty #7 > [ 1083.517127] Hardware name: Pine64 RockPro64 v2.1 (DT) > [ 1083.517574] Workqueue: events mmc_mq_recovery_handler > [ 1083.518032] pstate: 60000005 (nZCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--) > [ 1083.518645] pc : mmc_blk_reset+0x60/0x1ac > [ 1083.519004] lr : mmc_blk_reset+0x38/0x1ac > [ 1083.519361] sp : ffff8000100b3cd0 > [ 1083.519654] x29: ffff8000100b3cd0 x28: 0000000000000000 x27: 0000000000000000 > [ 1083.520288] x26: ffff80000b0ba000 x25: ffff0000f6e74805 x24: ffff000004c2fdc0 > [ 1083.520922] x23: ffff000014950000 x22: ffff000004c2fc18 x21: ffff00000a33c000 > [ 1083.521556] x20: 00000000ffffff85 x19: ffff000004c2fc00 x18: ffffffffffffffff > [ 1083.522189] x17: ffff80000cd9b200 x16: ffff80000cd9b190 x15: 0000000000000006 > [ 1083.522823] x14: 0000000000000000 x13: ffff80000b0c28f0 x12: 0000000000001707 > [ 1083.523457] x11: 00000000000007ad x10: ffff80000c6c28f0 x9 : ffff80000b0c28f0 > [ 1083.524090] x8 : 00000000fffbffff x7 : 0000000000000001 x6 : 0000000000000000 > [ 1083.524723] x5 : 0000000000000000 x4 : ffff0000f6e62d30 x3 : 0000000000000000 > [ 1083.525357] x2 : 0000000000000000 x1 : ffff00000b6e0000 x0 : 0000000000000000 > [ 1083.525990] Call trace: > [ 1083.526209] mmc_blk_reset+0x60/0x1ac > [ 1083.526536] mmc_blk_cqe_recovery+0x8c/0xd0 > [ 1083.526908] mmc_mq_recovery_handler+0xc4/0xd0 > [ 1083.527303] process_one_work+0x23c/0x3fc > [ 1083.527663] worker_thread+0x74/0x420 > [ 1083.527990] kthread+0xec/0xf0 > [ 1083.528264] ret_from_fork+0x10/0x20 > [ 1083.528587] Code: d50323bf d65f03c0 f94352a0 f9404000 (b9438c01) > [ 1083.529126] ---[ end trace 0000000000000000 ]--- > > [ 1431.677970] Unable to handle kernel NULL pointer dereference at virtual address 0000000000000000 > [ 1431.678753] Mem abort info: > [ 1431.679017] ESR = 0x96000004 > [ 1431.679303] EC = 0x25: DABT (current EL), IL = 32 bits > [ 1431.679786] SET = 0, FnV = 0 > [ 1431.680072] EA = 0, S1PTW = 0 > [ 1431.680366] FSC = 0x04: level 0 translation fault > [ 1431.680810] Data abort info: > [ 1431.681080] ISV = 0, ISS = 0x00000004 > [ 1431.681432] CM = 0, WnR = 0 > [ 1431.681712] user pgtable: 4k pages, 48-bit VAs, pgdp=000000000bb98000 > [ 1431.682390] [0000000000000000] pgd=0000000000000000, p4d=0000000000000000 > [ 1431.683393] Internal error: Oops: 96000004 [#1] SMP > [ 1431.683841] CPU: 0 PID: 19948 Comm: kworker/0:2 Not tainted 5.18.12-gf65532578f32-dirty #16 > [ 1431.684576] Hardware name: Pine64 RockPro64 v2.1 (DT) > [ 1431.685024] Workqueue: events mmc_mq_recovery_handler > [ 1431.685487] pstate: 60000005 (nZCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--) > [ 1431.686100] pc : mmc_put_card+0x38/0x110 > [ 1431.686453] lr : mmc_mq_recovery_handler+0x98/0xd0 > [ 1431.686879] sp : ffff800015813cf0 > [ 1431.687173] x29: ffff800015813cf0 x28: 0000000000000000 x27: 0000000000000000 > [ 1431.687807] x26: ffff80000b0ba000 x25: ffff0000f6e74805 x24: ffff000013bd65c0 > [ 1431.688441] x23: ffff000013b96120 x22: ffff000013bd6418 x21: 0000000000000000 > [ 1431.689075] x20: ffff800008ed1c70 x19: ffff8000091767d8 x18: ffffffffffffffff > [ 1431.689709] x17: 31335b1b6d375b1b x16: 6d305b1b47554245 x15: 0000000000000006 > [ 1431.690343] x14: 0000000000000000 x13: 0000000000000000 x12: 0000000000000000 > [ 1431.690976] x11: ffff000013bd6570 x10: 0000000000000001 x9 : ffff80000ea69228 > [ 1431.691611] x8 : ffff80000df892c8 x7 : 0000000000000000 x6 : 0000000000000001 > [ 1431.692245] x5 : 0000000000000001 x4 : 0000000000000002 x3 : ffff80000e6feac8 > [ 1431.692879] x2 : 000000000000036e x1 : ffff800008ed1c70 x0 : 0000000000000000 > [ 1431.693513] Call trace: > [ 1431.693732] mmc_put_card+0x38/0x110 > [ 1431.694055] mmc_mq_recovery_handler+0x98/0xd0 > [ 1431.694452] process_one_work+0x23c/0x3fc > [ 1431.694812] worker_thread+0x74/0x420 > [ 1431.695139] kthread+0xec/0xf0 > [ 1431.695414] ret_from_fork+0x10/0x20 > [ 1431.695738] Code: f9001bf7 aa0103f6 aa0003f5 aa1403e1 (f9400017) > [ 1431.696278] ---[ end trace 0000000000000000 ]--- > > Signed-off-by: Christian Loehle <cloehle@hyperstone.com> Thanks for finding these issues. A couple of comments below. > --- > drivers/mmc/core/block.c | 4 ++-- > drivers/mmc/core/queue.c | 5 +++-- > 2 files changed, 5 insertions(+), 4 deletions(-) > > diff --git a/drivers/mmc/core/block.c b/drivers/mmc/core/block.c > index ce89611a136e..0cd3a7065629 100644 > --- a/drivers/mmc/core/block.c > +++ b/drivers/mmc/core/block.c > @@ -997,8 +997,8 @@ static int mmc_blk_reset(struct mmc_blk_data *md, struct mmc_host *host, > > md->reset_done |= type; > err = mmc_hw_reset(host->card); > - /* Ensure we switch back to the correct partition */ > - if (err) { > + /* Ensure we switch back to the correct partition on successful reset */ > + if (!err) { This isn't quite right. Originally, this was err != -EOPNOTSUPP so "always" unless the reset was not attempted at all. When the -EOPNOTSUPP return value went away, this should have become unconditional. Also this change should be a separate patch, and have a fixes tag i.e. Fixes: fefdd3c91e0a ("mmc: core: Drop superfluous validations in mmc_hw|sw_reset()") > struct mmc_blk_data *main_md = > dev_get_drvdata(&host->card->dev); > int part_err; > diff --git a/drivers/mmc/core/queue.c b/drivers/mmc/core/queue.c > index fefaa901b50f..6931fa082ea7 100644 > --- a/drivers/mmc/core/queue.c > +++ b/drivers/mmc/core/queue.c > @@ -137,9 +137,10 @@ static void mmc_mq_recovery_handler(struct work_struct *work) > struct mmc_queue *mq = container_of(work, struct mmc_queue, > recovery_work); > struct request_queue *q = mq->queue; > + struct mmc_card *card = mq->card; > struct mmc_host *host = mq->card->host; > > - mmc_get_card(mq->card, &mq->ctx); > + mmc_get_card(card, &mq->ctx); > > mq->in_recovery = true; > > @@ -157,7 +158,7 @@ static void mmc_mq_recovery_handler(struct work_struct *work) > if (host->hsq_enabled) > host->cqe_ops->cqe_recovery_finish(host); > > - mmc_put_card(mq->card, &mq->ctx); > + mmc_put_card(card, &mq->ctx); > > blk_mq_run_hw_queues(q, true); > } Please try this instead: diff --git a/drivers/mmc/core/queue.c b/drivers/mmc/core/queue.c index 6931fa082ea7..d8d9115c51f6 100644 --- a/drivers/mmc/core/queue.c +++ b/drivers/mmc/core/queue.c @@ -494,6 +494,13 @@ void mmc_cleanup_queue(struct mmc_queue *mq) if (blk_queue_quiesced(q)) blk_mq_unquiesce_queue(q); + /* + * If the recovery completes the last (and only remaining) request in + * the queue, and the card has been removed, we could end up here with + * the recovery not quite finished yet, so flush it. + */ + flush_work(&mq->recovery_work); + blk_mq_free_tag_set(&mq->tag_set); /*
Thanks Adrian for the comments and hints, implemented and submitted. They also fix both issues. -----Original Message----- From: Adrian Hunter <adrian.hunter@intel.com> Sent: Mittwoch, 5. Oktober 2022 15:42 To: Christian Löhle <CLoehle@hyperstone.com>; ulf.hansson@linaro.org; Linux MMC List <linux-mmc@vger.kernel.org>; linux-kernel@vger.kernel.org Cc: Avri Altman <Avri.Altman@wdc.com> Subject: Re: [PATCHv2] mmc: core: fix race of queue reset and card removal On 4/10/22 16:13, Christian Löhle wrote: > If a recovery is active and the card is removed do not try to switch > back partitions. Furthermore do not reference > mq->card which might be NULLed in the meantime. > > This has been observed with recovery active with CQE. > [ 1083.510578] Unable to handle kernel NULL pointer dereference at > virtual address 000000000000038c [ 1083.511362] Mem abort info: > [ 1083.511626] ESR = 0x96000004 > [ 1083.511912] EC = 0x25: DABT (current EL), IL = 32 bits > [ 1083.512395] SET = 0, FnV = 0 > [ 1083.512681] EA = 0, S1PTW = 0 > [ 1083.512973] FSC = 0x04: level 0 translation fault > [ 1083.513417] Data abort info: > [ 1083.513686] ISV = 0, ISS = 0x00000004 > [ 1083.514039] CM = 0, WnR = 0 > [ 1083.514318] user pgtable: 4k pages, 48-bit VAs, > pgdp=000000000a4c3000 [ 1083.514899] [000000000000038c] > pgd=0000000000000000, p4d=0000000000000000 [ 1083.515854] Internal error: Oops: 96000004 [#1] SMP > [ 1083.516295] CPU: 0 PID: 153 Comm: kworker/0:2 Tainted: G W 5.18.12-g925ff1d10c99-dirty #7 > [ 1083.517127] Hardware name: Pine64 RockPro64 v2.1 (DT) [ > 1083.517574] Workqueue: events mmc_mq_recovery_handler [ 1083.518032] > pstate: 60000005 (nZCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--) [ > 1083.518645] pc : mmc_blk_reset+0x60/0x1ac [ 1083.519004] lr : > mmc_blk_reset+0x38/0x1ac [ 1083.519361] sp : ffff8000100b3cd0 [ > 1083.519654] x29: ffff8000100b3cd0 x28: 0000000000000000 x27: > 0000000000000000 [ 1083.520288] x26: ffff80000b0ba000 x25: > ffff0000f6e74805 x24: ffff000004c2fdc0 [ 1083.520922] x23: > ffff000014950000 x22: ffff000004c2fc18 x21: ffff00000a33c000 [ > 1083.521556] x20: 00000000ffffff85 x19: ffff000004c2fc00 x18: > ffffffffffffffff [ 1083.522189] x17: ffff80000cd9b200 x16: > ffff80000cd9b190 x15: 0000000000000006 [ 1083.522823] x14: > 0000000000000000 x13: ffff80000b0c28f0 x12: 0000000000001707 [ > 1083.523457] x11: 00000000000007ad x10: ffff80000c6c28f0 x9 : > ffff80000b0c28f0 [ 1083.524090] x8 : 00000000fffbffff x7 : > 0000000000000001 x6 : 0000000000000000 [ 1083.524723] x5 : 0000000000000000 x4 : ffff0000f6e62d30 x3 : 0000000000000000 [ 1083.525357] x2 : 0000000000000000 x1 : ffff00000b6e0000 x0 : 0000000000000000 [ 1083.525990] Call trace: > [ 1083.526209] mmc_blk_reset+0x60/0x1ac [ 1083.526536] > mmc_blk_cqe_recovery+0x8c/0xd0 [ 1083.526908] > mmc_mq_recovery_handler+0xc4/0xd0 [ 1083.527303] > process_one_work+0x23c/0x3fc [ 1083.527663] worker_thread+0x74/0x420 > [ 1083.527990] kthread+0xec/0xf0 [ 1083.528264] > ret_from_fork+0x10/0x20 [ 1083.528587] Code: d50323bf d65f03c0 > f94352a0 f9404000 (b9438c01) [ 1083.529126] ---[ end trace > 0000000000000000 ]--- > > [ 1431.677970] Unable to handle kernel NULL pointer dereference at > virtual address 0000000000000000 [ 1431.678753] Mem abort info: > [ 1431.679017] ESR = 0x96000004 > [ 1431.679303] EC = 0x25: DABT (current EL), IL = 32 bits > [ 1431.679786] SET = 0, FnV = 0 > [ 1431.680072] EA = 0, S1PTW = 0 > [ 1431.680366] FSC = 0x04: level 0 translation fault > [ 1431.680810] Data abort info: > [ 1431.681080] ISV = 0, ISS = 0x00000004 > [ 1431.681432] CM = 0, WnR = 0 > [ 1431.681712] user pgtable: 4k pages, 48-bit VAs, > pgdp=000000000bb98000 [ 1431.682390] [0000000000000000] > pgd=0000000000000000, p4d=0000000000000000 [ 1431.683393] Internal > error: Oops: 96000004 [#1] SMP [ 1431.683841] CPU: 0 PID: 19948 Comm: > kworker/0:2 Not tainted 5.18.12-gf65532578f32-dirty #16 [ 1431.684576] > Hardware name: Pine64 RockPro64 v2.1 (DT) [ 1431.685024] Workqueue: > events mmc_mq_recovery_handler [ 1431.685487] pstate: 60000005 (nZCv > daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--) [ 1431.686100] pc : > mmc_put_card+0x38/0x110 [ 1431.686453] lr : > mmc_mq_recovery_handler+0x98/0xd0 [ 1431.686879] sp : ffff800015813cf0 > [ 1431.687173] x29: ffff800015813cf0 x28: 0000000000000000 x27: > 0000000000000000 [ 1431.687807] x26: ffff80000b0ba000 x25: > ffff0000f6e74805 x24: ffff000013bd65c0 [ 1431.688441] x23: > ffff000013b96120 x22: ffff000013bd6418 x21: 0000000000000000 [ > 1431.689075] x20: ffff800008ed1c70 x19: ffff8000091767d8 x18: > ffffffffffffffff [ 1431.689709] x17: 31335b1b6d375b1b x16: > 6d305b1b47554245 x15: 0000000000000006 [ 1431.690343] x14: > 0000000000000000 x13: 0000000000000000 x12: 0000000000000000 [ > 1431.690976] x11: ffff000013bd6570 x10: 0000000000000001 x9 : > ffff80000ea69228 [ 1431.691611] x8 : ffff80000df892c8 x7 : > 0000000000000000 x6 : 0000000000000001 [ 1431.692245] x5 : 0000000000000001 x4 : 0000000000000002 x3 : ffff80000e6feac8 [ 1431.692879] x2 : 000000000000036e x1 : ffff800008ed1c70 x0 : 0000000000000000 [ 1431.693513] Call trace: > [ 1431.693732] mmc_put_card+0x38/0x110 [ 1431.694055] > mmc_mq_recovery_handler+0x98/0xd0 [ 1431.694452] > process_one_work+0x23c/0x3fc [ 1431.694812] worker_thread+0x74/0x420 > [ 1431.695139] kthread+0xec/0xf0 [ 1431.695414] > ret_from_fork+0x10/0x20 [ 1431.695738] Code: f9001bf7 aa0103f6 > aa0003f5 aa1403e1 (f9400017) [ 1431.696278] ---[ end trace > 0000000000000000 ]--- > > Signed-off-by: Christian Loehle <cloehle@hyperstone.com> Thanks for finding these issues. A couple of comments below. > --- > drivers/mmc/core/block.c | 4 ++-- > drivers/mmc/core/queue.c | 5 +++-- > 2 files changed, 5 insertions(+), 4 deletions(-) > > diff --git a/drivers/mmc/core/block.c b/drivers/mmc/core/block.c index > ce89611a136e..0cd3a7065629 100644 > --- a/drivers/mmc/core/block.c > +++ b/drivers/mmc/core/block.c > @@ -997,8 +997,8 @@ static int mmc_blk_reset(struct mmc_blk_data *md, > struct mmc_host *host, > > md->reset_done |= type; > err = mmc_hw_reset(host->card); > - /* Ensure we switch back to the correct partition */ > - if (err) { > + /* Ensure we switch back to the correct partition on successful reset */ > + if (!err) { This isn't quite right. Originally, this was err != -EOPNOTSUPP so "always" unless the reset was not attempted at all. When the -EOPNOTSUPP return value went away, this should have become unconditional. Also this change should be a separate patch, and have a fixes tag i.e. Fixes: fefdd3c91e0a ("mmc: core: Drop superfluous validations in mmc_hw|sw_reset()") > struct mmc_blk_data *main_md = > dev_get_drvdata(&host->card->dev); > int part_err; > diff --git a/drivers/mmc/core/queue.c b/drivers/mmc/core/queue.c index > fefaa901b50f..6931fa082ea7 100644 > --- a/drivers/mmc/core/queue.c > +++ b/drivers/mmc/core/queue.c > @@ -137,9 +137,10 @@ static void mmc_mq_recovery_handler(struct work_struct *work) > struct mmc_queue *mq = container_of(work, struct mmc_queue, > recovery_work); > struct request_queue *q = mq->queue; > + struct mmc_card *card = mq->card; > struct mmc_host *host = mq->card->host; > > - mmc_get_card(mq->card, &mq->ctx); > + mmc_get_card(card, &mq->ctx); > > mq->in_recovery = true; > > @@ -157,7 +158,7 @@ static void mmc_mq_recovery_handler(struct work_struct *work) > if (host->hsq_enabled) > host->cqe_ops->cqe_recovery_finish(host); > > - mmc_put_card(mq->card, &mq->ctx); > + mmc_put_card(card, &mq->ctx); > > blk_mq_run_hw_queues(q, true); > } Please try this instead: diff --git a/drivers/mmc/core/queue.c b/drivers/mmc/core/queue.c index 6931fa082ea7..d8d9115c51f6 100644 --- a/drivers/mmc/core/queue.c +++ b/drivers/mmc/core/queue.c @@ -494,6 +494,13 @@ void mmc_cleanup_queue(struct mmc_queue *mq) if (blk_queue_quiesced(q)) blk_mq_unquiesce_queue(q); + /* + * If the recovery completes the last (and only remaining) request in + * the queue, and the card has been removed, we could end up here with + * the recovery not quite finished yet, so flush it. + */ + flush_work(&mq->recovery_work); + blk_mq_free_tag_set(&mq->tag_set); /* Hyperstone GmbH | Reichenaustr. 39a | 78467 Konstanz Managing Director: Dr. Jan Peter Berns. Commercial register of local courts: Freiburg HRB381782
diff --git a/drivers/mmc/core/block.c b/drivers/mmc/core/block.c index ce89611a136e..0cd3a7065629 100644 --- a/drivers/mmc/core/block.c +++ b/drivers/mmc/core/block.c @@ -997,8 +997,8 @@ static int mmc_blk_reset(struct mmc_blk_data *md, struct mmc_host *host, md->reset_done |= type; err = mmc_hw_reset(host->card); - /* Ensure we switch back to the correct partition */ - if (err) { + /* Ensure we switch back to the correct partition on successful reset */ + if (!err) { struct mmc_blk_data *main_md = dev_get_drvdata(&host->card->dev); int part_err; diff --git a/drivers/mmc/core/queue.c b/drivers/mmc/core/queue.c index fefaa901b50f..6931fa082ea7 100644 --- a/drivers/mmc/core/queue.c +++ b/drivers/mmc/core/queue.c @@ -137,9 +137,10 @@ static void mmc_mq_recovery_handler(struct work_struct *work) struct mmc_queue *mq = container_of(work, struct mmc_queue, recovery_work); struct request_queue *q = mq->queue; + struct mmc_card *card = mq->card; struct mmc_host *host = mq->card->host; - mmc_get_card(mq->card, &mq->ctx); + mmc_get_card(card, &mq->ctx); mq->in_recovery = true; @@ -157,7 +158,7 @@ static void mmc_mq_recovery_handler(struct work_struct *work) if (host->hsq_enabled) host->cqe_ops->cqe_recovery_finish(host); - mmc_put_card(mq->card, &mq->ctx); + mmc_put_card(card, &mq->ctx); blk_mq_run_hw_queues(q, true); }
If a recovery is active and the card is removed do not try to switch back partitions. Furthermore do not reference mq->card which might be NULLed in the meantime. This has been observed with recovery active with CQE. [ 1083.510578] Unable to handle kernel NULL pointer dereference at virtual address 000000000000038c [ 1083.511362] Mem abort info: [ 1083.511626] ESR = 0x96000004 [ 1083.511912] EC = 0x25: DABT (current EL), IL = 32 bits [ 1083.512395] SET = 0, FnV = 0 [ 1083.512681] EA = 0, S1PTW = 0 [ 1083.512973] FSC = 0x04: level 0 translation fault [ 1083.513417] Data abort info: [ 1083.513686] ISV = 0, ISS = 0x00000004 [ 1083.514039] CM = 0, WnR = 0 [ 1083.514318] user pgtable: 4k pages, 48-bit VAs, pgdp=000000000a4c3000 [ 1083.514899] [000000000000038c] pgd=0000000000000000, p4d=0000000000000000 [ 1083.515854] Internal error: Oops: 96000004 [#1] SMP [ 1083.516295] CPU: 0 PID: 153 Comm: kworker/0:2 Tainted: G W 5.18.12-g925ff1d10c99-dirty #7 [ 1083.517127] Hardware name: Pine64 RockPro64 v2.1 (DT) [ 1083.517574] Workqueue: events mmc_mq_recovery_handler [ 1083.518032] pstate: 60000005 (nZCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--) [ 1083.518645] pc : mmc_blk_reset+0x60/0x1ac [ 1083.519004] lr : mmc_blk_reset+0x38/0x1ac [ 1083.519361] sp : ffff8000100b3cd0 [ 1083.519654] x29: ffff8000100b3cd0 x28: 0000000000000000 x27: 0000000000000000 [ 1083.520288] x26: ffff80000b0ba000 x25: ffff0000f6e74805 x24: ffff000004c2fdc0 [ 1083.520922] x23: ffff000014950000 x22: ffff000004c2fc18 x21: ffff00000a33c000 [ 1083.521556] x20: 00000000ffffff85 x19: ffff000004c2fc00 x18: ffffffffffffffff [ 1083.522189] x17: ffff80000cd9b200 x16: ffff80000cd9b190 x15: 0000000000000006 [ 1083.522823] x14: 0000000000000000 x13: ffff80000b0c28f0 x12: 0000000000001707 [ 1083.523457] x11: 00000000000007ad x10: ffff80000c6c28f0 x9 : ffff80000b0c28f0 [ 1083.524090] x8 : 00000000fffbffff x7 : 0000000000000001 x6 : 0000000000000000 [ 1083.524723] x5 : 0000000000000000 x4 : ffff0000f6e62d30 x3 : 0000000000000000 [ 1083.525357] x2 : 0000000000000000 x1 : ffff00000b6e0000 x0 : 0000000000000000 [ 1083.525990] Call trace: [ 1083.526209] mmc_blk_reset+0x60/0x1ac [ 1083.526536] mmc_blk_cqe_recovery+0x8c/0xd0 [ 1083.526908] mmc_mq_recovery_handler+0xc4/0xd0 [ 1083.527303] process_one_work+0x23c/0x3fc [ 1083.527663] worker_thread+0x74/0x420 [ 1083.527990] kthread+0xec/0xf0 [ 1083.528264] ret_from_fork+0x10/0x20 [ 1083.528587] Code: d50323bf d65f03c0 f94352a0 f9404000 (b9438c01) [ 1083.529126] ---[ end trace 0000000000000000 ]--- [ 1431.677970] Unable to handle kernel NULL pointer dereference at virtual address 0000000000000000 [ 1431.678753] Mem abort info: [ 1431.679017] ESR = 0x96000004 [ 1431.679303] EC = 0x25: DABT (current EL), IL = 32 bits [ 1431.679786] SET = 0, FnV = 0 [ 1431.680072] EA = 0, S1PTW = 0 [ 1431.680366] FSC = 0x04: level 0 translation fault [ 1431.680810] Data abort info: [ 1431.681080] ISV = 0, ISS = 0x00000004 [ 1431.681432] CM = 0, WnR = 0 [ 1431.681712] user pgtable: 4k pages, 48-bit VAs, pgdp=000000000bb98000 [ 1431.682390] [0000000000000000] pgd=0000000000000000, p4d=0000000000000000 [ 1431.683393] Internal error: Oops: 96000004 [#1] SMP [ 1431.683841] CPU: 0 PID: 19948 Comm: kworker/0:2 Not tainted 5.18.12-gf65532578f32-dirty #16 [ 1431.684576] Hardware name: Pine64 RockPro64 v2.1 (DT) [ 1431.685024] Workqueue: events mmc_mq_recovery_handler [ 1431.685487] pstate: 60000005 (nZCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--) [ 1431.686100] pc : mmc_put_card+0x38/0x110 [ 1431.686453] lr : mmc_mq_recovery_handler+0x98/0xd0 [ 1431.686879] sp : ffff800015813cf0 [ 1431.687173] x29: ffff800015813cf0 x28: 0000000000000000 x27: 0000000000000000 [ 1431.687807] x26: ffff80000b0ba000 x25: ffff0000f6e74805 x24: ffff000013bd65c0 [ 1431.688441] x23: ffff000013b96120 x22: ffff000013bd6418 x21: 0000000000000000 [ 1431.689075] x20: ffff800008ed1c70 x19: ffff8000091767d8 x18: ffffffffffffffff [ 1431.689709] x17: 31335b1b6d375b1b x16: 6d305b1b47554245 x15: 0000000000000006 [ 1431.690343] x14: 0000000000000000 x13: 0000000000000000 x12: 0000000000000000 [ 1431.690976] x11: ffff000013bd6570 x10: 0000000000000001 x9 : ffff80000ea69228 [ 1431.691611] x8 : ffff80000df892c8 x7 : 0000000000000000 x6 : 0000000000000001 [ 1431.692245] x5 : 0000000000000001 x4 : 0000000000000002 x3 : ffff80000e6feac8 [ 1431.692879] x2 : 000000000000036e x1 : ffff800008ed1c70 x0 : 0000000000000000 [ 1431.693513] Call trace: [ 1431.693732] mmc_put_card+0x38/0x110 [ 1431.694055] mmc_mq_recovery_handler+0x98/0xd0 [ 1431.694452] process_one_work+0x23c/0x3fc [ 1431.694812] worker_thread+0x74/0x420 [ 1431.695139] kthread+0xec/0xf0 [ 1431.695414] ret_from_fork+0x10/0x20 [ 1431.695738] Code: f9001bf7 aa0103f6 aa0003f5 aa1403e1 (f9400017) [ 1431.696278] ---[ end trace 0000000000000000 ]--- Signed-off-by: Christian Loehle <cloehle@hyperstone.com> --- drivers/mmc/core/block.c | 4 ++-- drivers/mmc/core/queue.c | 5 +++-- 2 files changed, 5 insertions(+), 4 deletions(-)