December 2017 - Linux-stable-mirror

[Linux-stable-mirror] FAILED: patch "[PATCH] drm/amdgpu: Use unsigned ring indices in amdgpu_queue_mgr_map" failed to apply to 4.14-stable tree

by gregkh＠linuxfoundation.org

The patch below does not apply to the 4.14-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable(a)vger.kernel.org>. thanks, greg k-h ------------------ original commit in Linus's tree ------------------ >From fa7c7939b4bf112cd06ba166b739244077898990 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= <michel.daenzer(a)amd.com> Date: Wed, 22 Nov 2017 15:55:21 +0100 Subject: [PATCH] drm/amdgpu: Use unsigned ring indices in amdgpu_queue_mgr_map MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This matches the corresponding UAPI fields. Treating the ring index as signed could result in accessing random unrelated memory if the MSB was set. Fixes: effd924d2f3b ("drm/amdgpu: untie user ring ids from kernel ring ids v6") Cc: stable(a)vger.kernel.org Reviewed-by: Alex Deucher <alexander.deucher(a)amd.com> Reviewed-by: Christian König <christian.koenig(a)amd.com> Signed-off-by: Michel Dänzer <michel.daenzer(a)amd.com> Signed-off-by: Alex Deucher <alexander.deucher(a)amd.com> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index c25cedff4915..0b14b5373783 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -717,7 +717,7 @@ int amdgpu_queue_mgr_fini(struct amdgpu_device *adev, struct amdgpu_queue_mgr *mgr); int amdgpu_queue_mgr_map(struct amdgpu_device *adev, struct amdgpu_queue_mgr *mgr, - int hw_ip, int instance, int ring, + u32 hw_ip, u32 instance, u32 ring, struct amdgpu_ring **out_ring); /* diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_queue_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_queue_mgr.c index 190e28cb827e..93d86619e802 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_queue_mgr.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_queue_mgr.c @@ -63,7 +63,7 @@ static int amdgpu_update_cached_map(struct amdgpu_queue_mapper *mapper, static int amdgpu_identity_map(struct amdgpu_device *adev, struct amdgpu_queue_mapper *mapper, - int ring, + u32 ring, struct amdgpu_ring **out_ring) { switch (mapper->hw_ip) { @@ -121,7 +121,7 @@ static enum amdgpu_ring_type amdgpu_hw_ip_to_ring_type(int hw_ip) static int amdgpu_lru_map(struct amdgpu_device *adev, struct amdgpu_queue_mapper *mapper, - int user_ring, bool lru_pipe_order, + u32 user_ring, bool lru_pipe_order, struct amdgpu_ring **out_ring) { int r, i, j; @@ -208,7 +208,7 @@ int amdgpu_queue_mgr_fini(struct amdgpu_device *adev, */ int amdgpu_queue_mgr_map(struct amdgpu_device *adev, struct amdgpu_queue_mgr *mgr, - int hw_ip, int instance, int ring, + u32 hw_ip, u32 instance, u32 ring, struct amdgpu_ring **out_ring) { int r, ip_num_rings;

7 years, 1 month

3
2
0 0

[Linux-stable-mirror] FAILED: patch "[PATCH] mm, oom_reaper: gather each vma to prevent leaking TLB entry" failed to apply to 4.9-stable tree

by gregkh＠linuxfoundation.org

The patch below does not apply to the 4.9-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable(a)vger.kernel.org>. thanks, greg k-h ------------------ original commit in Linus's tree ------------------ >From 687cb0884a714ff484d038e9190edc874edcf146 Mon Sep 17 00:00:00 2001 From: Wang Nan <wangnan0(a)huawei.com> Date: Wed, 29 Nov 2017 16:09:58 -0800 Subject: [PATCH] mm, oom_reaper: gather each vma to prevent leaking TLB entry tlb_gather_mmu(&tlb, mm, 0, -1) means gathering the whole virtual memory space. In this case, tlb->fullmm is true. Some archs like arm64 doesn't flush TLB when tlb->fullmm is true: commit 5a7862e83000 ("arm64: tlbflush: avoid flushing when fullmm == 1"). Which causes leaking of tlb entries. Will clarifies his patch: "Basically, we tag each address space with an ASID (PCID on x86) which is resident in the TLB. This means we can elide TLB invalidation when pulling down a full mm because we won't ever assign that ASID to another mm without doing TLB invalidation elsewhere (which actually just nukes the whole TLB). I think that means that we could potentially not fault on a kernel uaccess, because we could hit in the TLB" There could be a window between complete_signal() sending IPI to other cores and all threads sharing this mm are really kicked off from cores. In this window, the oom reaper may calls tlb_flush_mmu_tlbonly() to flush TLB then frees pages. However, due to the above problem, the TLB entries are not really flushed on arm64. Other threads are possible to access these pages through TLB entries. Moreover, a copy_to_user() can also write to these pages without generating page fault, causes use-after-free bugs. This patch gathers each vma instead of gathering full vm space. In this case tlb->fullmm is not true. The behavior of oom reaper become similar to munmapping before do_exit, which should be safe for all archs. Link: http://lkml.kernel.org/r/20171107095453.179940-1-wangnan0@huawei.com Fixes: aac453635549 ("mm, oom: introduce oom reaper") Signed-off-by: Wang Nan <wangnan0(a)huawei.com> Acked-by: Michal Hocko <mhocko(a)suse.com> Acked-by: David Rientjes <rientjes(a)google.com> Cc: Minchan Kim <minchan(a)kernel.org> Cc: Will Deacon <will.deacon(a)arm.com> Cc: Bob Liu <liubo95(a)huawei.com> Cc: Ingo Molnar <mingo(a)kernel.org> Cc: Roman Gushchin <guro(a)fb.com> Cc: Konstantin Khlebnikov <khlebnikov(a)yandex-team.ru> Cc: Andrea Arcangeli <aarcange(a)redhat.com> Cc: <stable(a)vger.kernel.org> Signed-off-by: Andrew Morton <akpm(a)linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds(a)linux-foundation.org> diff --git a/mm/oom_kill.c b/mm/oom_kill.c index c86fbd1b590e..c957be32b27a 100644 --- a/mm/oom_kill.c +++ b/mm/oom_kill.c @@ -550,7 +550,6 @@ static bool __oom_reap_task_mm(struct task_struct *tsk, struct mm_struct *mm) */ set_bit(MMF_UNSTABLE, &mm->flags); - tlb_gather_mmu(&tlb, mm, 0, -1); for (vma = mm->mmap ; vma; vma = vma->vm_next) { if (!can_madv_dontneed_vma(vma)) continue; @@ -565,11 +564,13 @@ static bool __oom_reap_task_mm(struct task_struct *tsk, struct mm_struct *mm) * we do not want to block exit_mmap by keeping mm ref * count elevated without a good reason. */ - if (vma_is_anonymous(vma) || !(vma->vm_flags & VM_SHARED)) + if (vma_is_anonymous(vma) || !(vma->vm_flags & VM_SHARED)) { + tlb_gather_mmu(&tlb, mm, vma->vm_start, vma->vm_end); unmap_page_range(&tlb, vma, vma->vm_start, vma->vm_end, NULL); + tlb_finish_mmu(&tlb, vma->vm_start, vma->vm_end); + } } - tlb_finish_mmu(&tlb, 0, -1); pr_info("oom_reaper: reaped process %d (%s), now anon-rss:%lukB, file-rss:%lukB, shmem-rss:%lukB\n", task_pid_nr(tsk), tsk->comm, K(get_mm_counter(mm, MM_ANONPAGES)),

7 years, 1 month

3
2
0 0

[Linux-stable-mirror] v4.9.67 build: 0 failures 0 warnings (v4.9.67)

by Build bot for Mark Brown

Tree/Branch: v4.9.67 Git describe: v4.9.67 Commit: 284bbc7824 Linux 4.9.67 Build Time: 101 min 31 sec Passed: 10 / 10 (100.00 %) Failed: 0 / 10 ( 0.00 %) Errors: 0 Warnings: 0 Section Mismatches: 0 ------------------------------------------------------------------------------- defconfigs with issues (other than build errors): ------------------------------------------------------------------------------- =============================================================================== Detailed per-defconfig build reports below: ------------------------------------------------------------------------------- Passed with no errors, warnings or mismatches: arm64-allnoconfig arm64-allmodconfig arm-multi_v5_defconfig arm-multi_v7_defconfig x86_64-defconfig arm-allmodconfig arm-allnoconfig x86_64-allnoconfig arm-multi_v4t_defconfig arm64-defconfig close failed in file object destructor: sys.excepthook is missing lost sys.stderr

7 years, 1 month

1
0
0 0

[Linux-stable-mirror] [PATCH v3 1/2] Ensure that the SCSI error handler gets woken up

by Bart Van Assche

If scsi_eh_scmd_add() is called concurrently with scsi_host_queue_ready() while shost->host_blocked > 0 then it can happen that neither function wakes up the SCSI error handler. Fix this by making every function that decreases the host_busy counter wake up the error handler if necessary and by protecting the host_failed checks with the SCSI host lock. Reported-by: Pavel Tikhomirov <ptikhomirov(a)virtuozzo.com> References: https://marc.info/?l=linux-kernel&m=150461610630736 Fixes: commit 746650160866 ("scsi: convert host_busy to atomic_t") Signed-off-by: Bart Van Assche <bart.vanassche(a)wdc.com> Cc: Konstantin Khorenko <khorenko(a)virtuozzo.com> Cc: Stuart Hayes <stuart.w.hayes(a)gmail.com> Cc: Pavel Tikhomirov <ptikhomirov(a)virtuozzo.com> Cc: Christoph Hellwig <hch(a)lst.de> Cc: Hannes Reinecke <hare(a)suse.com> Cc: Johannes Thumshirn <jthumshirn(a)suse.de> Cc: <stable(a)vger.kernel.org> --- drivers/scsi/hosts.c | 6 ++++++ drivers/scsi/scsi_error.c | 18 ++++++++++++++++-- drivers/scsi/scsi_lib.c | 39 ++++++++++++++++++++++++++++----------- include/scsi/scsi_host.h | 2 ++ 4 files changed, 52 insertions(+), 13 deletions(-) diff --git a/drivers/scsi/hosts.c b/drivers/scsi/hosts.c index a306af6a5ea7..a0a7e4ff255c 100644 --- a/drivers/scsi/hosts.c +++ b/drivers/scsi/hosts.c @@ -324,6 +324,9 @@ static void scsi_host_dev_release(struct device *dev) scsi_proc_hostdir_rm(shost->hostt); + /* Wait for functions invoked through call_rcu(&shost->rcu, ...) */ + rcu_barrier(); + if (shost->tmf_work_q) destroy_workqueue(shost->tmf_work_q); if (shost->ehandler) @@ -331,6 +334,8 @@ static void scsi_host_dev_release(struct device *dev) if (shost->work_q) destroy_workqueue(shost->work_q); + destroy_rcu_head(&shost->rcu); + if (shost->shost_state == SHOST_CREATED) { /* * Free the shost_dev device name here if scsi_host_alloc() @@ -399,6 +404,7 @@ struct Scsi_Host *scsi_host_alloc(struct scsi_host_template *sht, int privsize) INIT_LIST_HEAD(&shost->starved_list); init_waitqueue_head(&shost->host_wait); mutex_init(&shost->scan_mutex); + init_rcu_head(&shost->rcu); index = ida_simple_get(&host_index_ida, 0, 0, GFP_KERNEL); if (index < 0) diff --git a/drivers/scsi/scsi_error.c b/drivers/scsi/scsi_error.c index 5e89049e9b4e..258b8a741992 100644 --- a/drivers/scsi/scsi_error.c +++ b/drivers/scsi/scsi_error.c @@ -226,6 +226,17 @@ static void scsi_eh_reset(struct scsi_cmnd *scmd) } } +static void scsi_eh_inc_host_failed(struct rcu_head *head) +{ + struct Scsi_Host *shost = container_of(head, typeof(*shost), rcu); + unsigned long flags; + + spin_lock_irqsave(shost->host_lock, flags); + shost->host_failed++; + scsi_eh_wakeup(shost); + spin_unlock_irqrestore(shost->host_lock, flags); +} + /** * scsi_eh_scmd_add - add scsi cmd to error handling. * @scmd: scmd to run eh on. @@ -248,9 +259,12 @@ void scsi_eh_scmd_add(struct scsi_cmnd *scmd) scsi_eh_reset(scmd); list_add_tail(&scmd->eh_entry, &shost->eh_cmd_q); - shost->host_failed++; - scsi_eh_wakeup(shost); spin_unlock_irqrestore(shost->host_lock, flags); + /* + * Ensure that all tasks observe the host state change before the + * host_failed change. + */ + call_rcu(&shost->rcu, scsi_eh_inc_host_failed); } /** diff --git a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c index b6d3842b6809..5cbc69b2b1ae 100644 --- a/drivers/scsi/scsi_lib.c +++ b/drivers/scsi/scsi_lib.c @@ -318,22 +318,39 @@ static void scsi_init_cmd_errh(struct scsi_cmnd *cmd) cmd->cmd_len = scsi_command_size(cmd->cmnd); } -void scsi_device_unbusy(struct scsi_device *sdev) +/* + * Decrement the host_busy counter and wake up the error handler if necessary. + * Avoid as follows that the error handler is not woken up if shost->host_busy + * == shost->host_failed: use call_rcu() in scsi_eh_scmd_add() in combination + * with an RCU read lock in this function to ensure that this function in its + * entirety either finishes before scsi_eh_scmd_add() increases the + * host_failed counter or that it notices the shost state change made by + * scsi_eh_scmd_add(). + */ +static void scsi_dec_host_busy(struct Scsi_Host *shost) { - struct Scsi_Host *shost = sdev->host; - struct scsi_target *starget = scsi_target(sdev); unsigned long flags; + rcu_read_lock(); atomic_dec(&shost->host_busy); - if (starget->can_queue > 0) - atomic_dec(&starget->target_busy); - - if (unlikely(scsi_host_in_recovery(shost) && - (shost->host_failed || shost->host_eh_scheduled))) { + if (unlikely(scsi_host_in_recovery(shost))) { spin_lock_irqsave(shost->host_lock, flags); - scsi_eh_wakeup(shost); + if (shost->host_failed || shost->host_eh_scheduled) + scsi_eh_wakeup(shost); spin_unlock_irqrestore(shost->host_lock, flags); } + rcu_read_unlock(); +} + +void scsi_device_unbusy(struct scsi_device *sdev) +{ + struct Scsi_Host *shost = sdev->host; + struct scsi_target *starget = scsi_target(sdev); + + scsi_dec_host_busy(shost); + + if (starget->can_queue > 0) + atomic_dec(&starget->target_busy); atomic_dec(&sdev->device_busy); } @@ -1531,7 +1548,7 @@ static inline int scsi_host_queue_ready(struct request_queue *q, list_add_tail(&sdev->starved_entry, &shost->starved_list); spin_unlock_irq(shost->host_lock); out_dec: - atomic_dec(&shost->host_busy); + scsi_dec_host_busy(shost); return 0; } @@ -2017,7 +2034,7 @@ static blk_status_t scsi_queue_rq(struct blk_mq_hw_ctx *hctx, return BLK_STS_OK; out_dec_host_busy: - atomic_dec(&shost->host_busy); + scsi_dec_host_busy(shost); out_dec_target_busy: if (scsi_target(sdev)->can_queue > 0) atomic_dec(&scsi_target(sdev)->target_busy); diff --git a/include/scsi/scsi_host.h b/include/scsi/scsi_host.h index a8b7bf879ced..1a1df0d21ee3 100644 --- a/include/scsi/scsi_host.h +++ b/include/scsi/scsi_host.h @@ -571,6 +571,8 @@ struct Scsi_Host { struct blk_mq_tag_set tag_set; }; + struct rcu_head rcu; + atomic_t host_busy; /* commands actually active on low-level */ atomic_t host_blocked; -- 2.15.0

7 years, 1 month

3
2
0 0

Re: [Linux-stable-mirror] UFS utilities

by Bean Huo (beanhuo)

Hi, greg k-h > >So what UFS commands are you missing that you need to see implemented? > >And again, have you checked the different forks of the driver? > Seems there is something misunderstood, I want to use UPIU, rather than CDB. Maybe it is not possible based on current UFS stacks. Of course, exactly, there is no missing SCSI command listed in UFS 2.1. >> >> And also it doesn't support several UFS special command. >> > >> >Are you referring to SCSI commands or rather to UFS commands that >> >fall outside the SCSI spec? Anyway, an approach that is used by many >> >SCSI drivers to export information to user space that falls outside >> >the SCSI spec is to create additional sysfs attributes. See also the >> >sdev_attrs and shost_attrs members of struct scsi_host_template. >> > >> Yes, for the UFS information, I can use these interface/approach to easily >get. >> I am thinking how about some testing case and configuration operation. > >Which ones exactly? > >> Also, is it possible bypass SCSI stacks and go into directly UFS stack? > >Look at the different sysfs files for the UFS device, it does that for some >commands. > To be honest, I don't know which interface, it can pass UPIU to UFS driver, And bypass SCSI stacks. Thanks. Bean Huo

7 years, 1 month

2
1
0 0

[Linux-stable-mirror] [PATCH 4.9-stable] Revert "crypto: caam - get rid of tasklet"

by Horia Geantă

commit 2b163b5bce04546da72617bfb6c8bf07a45c4b17 upstream. This reverts commit 66d2e2028091a074aa1290d2eeda5ddb1a6c329c. Quoting from Russell's findings: https://www.mail-archive.com/linux-crypto@vger.kernel.org/msg21136.html [quote] Okay, I've re-tested, using a different way of measuring, because using openssl speed is impractical for off-loaded engines. I've decided to use this way to measure the performance: dd if=/dev/zero bs=1048576 count=128 | /usr/bin/time openssl dgst -md5 For the threaded IRQs case gives: 0.05user 2.74system 0:05.30elapsed 52%CPU (0avgtext+0avgdata 2400maxresident)k 0.06user 2.52system 0:05.18elapsed 49%CPU (0avgtext+0avgdata 2404maxresident)k 0.12user 2.60system 0:05.61elapsed 48%CPU (0avgtext+0avgdata 2460maxresident)k => 5.36s => 25.0MB/s and the tasklet case: 0.08user 2.53system 0:04.83elapsed 54%CPU (0avgtext+0avgdata 2468maxresident)k 0.09user 2.47system 0:05.16elapsed 49%CPU (0avgtext+0avgdata 2368maxresident)k 0.10user 2.51system 0:04.87elapsed 53%CPU (0avgtext+0avgdata 2460maxresident)k => 4.95 => 27.1MB/s which corresponds to an 8% slowdown for the threaded IRQ case. So, tasklets are indeed faster than threaded IRQs. [...] I think I've proven from the above that this patch needs to be reverted due to the performance regression, and that there _is_ most definitely a deterimental effect of switching from tasklets to threaded IRQs. [/quote] Signed-off-by: Horia Geantă <horia.geanta(a)nxp.com> Signed-off-by: Herbert Xu <herbert(a)gondor.apana.org.au> --- Mihai Ordean reported soft lockups at IPsec ESP high rates on i.MX6Q, on kernels 4.9.{35,36}. This patch, cherry-picked from 4.10, fixes the issue. drivers/crypto/caam/intern.h | 1 + drivers/crypto/caam/jr.c | 25 ++++++++++++++++--------- 2 files changed, 17 insertions(+), 9 deletions(-) diff --git a/drivers/crypto/caam/intern.h b/drivers/crypto/caam/intern.h index 5d4c05074a5c..e2bcacc1a921 100644 --- a/drivers/crypto/caam/intern.h +++ b/drivers/crypto/caam/intern.h @@ -41,6 +41,7 @@ struct caam_drv_private_jr { struct device *dev; int ridx; struct caam_job_ring __iomem *rregs; /* JobR's register space */ + struct tasklet_struct irqtask; int irq; /* One per queue */ /* Number of scatterlist crypt transforms active on the JobR */ diff --git a/drivers/crypto/caam/jr.c b/drivers/crypto/caam/jr.c index 757c27f9953d..9e7f28122bb7 100644 --- a/drivers/crypto/caam/jr.c +++ b/drivers/crypto/caam/jr.c @@ -73,6 +73,8 @@ static int caam_jr_shutdown(struct device *dev) ret = caam_reset_hw_jr(dev); + tasklet_kill(&jrp->irqtask); + /* Release interrupt */ free_irq(jrp->irq, dev); @@ -128,7 +130,7 @@ static irqreturn_t caam_jr_interrupt(int irq, void *st_dev) /* * Check the output ring for ready responses, kick - * the threaded irq if jobs done. + * tasklet if jobs done. */ irqstate = rd_reg32(&jrp->rregs->jrintstatus); if (!irqstate) @@ -150,13 +152,18 @@ static irqreturn_t caam_jr_interrupt(int irq, void *st_dev) /* Have valid interrupt at this point, just ACK and trigger */ wr_reg32(&jrp->rregs->jrintstatus, irqstate); - return IRQ_WAKE_THREAD; + preempt_disable(); + tasklet_schedule(&jrp->irqtask); + preempt_enable(); + + return IRQ_HANDLED; } -static irqreturn_t caam_jr_threadirq(int irq, void *st_dev) +/* Deferred service handler, run as interrupt-fired tasklet */ +static void caam_jr_dequeue(unsigned long devarg) { int hw_idx, sw_idx, i, head, tail; - struct device *dev = st_dev; + struct device *dev = (struct device *)devarg; struct caam_drv_private_jr *jrp = dev_get_drvdata(dev); void (*usercall)(struct device *dev, u32 *desc, u32 status, void *arg); u32 *userdesc, userstatus; @@ -230,8 +237,6 @@ static irqreturn_t caam_jr_threadirq(int irq, void *st_dev) /* reenable / unmask IRQs */ clrsetbits_32(&jrp->rregs->rconfig_lo, JRCFG_IMSK, 0); - - return IRQ_HANDLED; } /** @@ -389,10 +394,11 @@ static int caam_jr_init(struct device *dev) jrp = dev_get_drvdata(dev); + tasklet_init(&jrp->irqtask, caam_jr_dequeue, (unsigned long)dev); + /* Connect job ring interrupt handler. */ - error = request_threaded_irq(jrp->irq, caam_jr_interrupt, - caam_jr_threadirq, IRQF_SHARED, - dev_name(dev), dev); + error = request_irq(jrp->irq, caam_jr_interrupt, IRQF_SHARED, + dev_name(dev), dev); if (error) { dev_err(dev, "can't connect JobR %d interrupt (%d)\n", jrp->ridx, jrp->irq); @@ -454,6 +460,7 @@ static int caam_jr_init(struct device *dev) out_free_irq: free_irq(jrp->irq, dev); out_kill_deq: + tasklet_kill(&jrp->irqtask); return error; } -- 2.12.0.264.gd6db3f216544

7 years, 1 month

2
1
0 0

[Linux-stable-mirror] Patch "Revert "crypto: caam - get rid of tasklet"" has been added to the 4.9-stable tree

by gregkh＠linuxfoundation.org

This is a note to let you know that I've just added the patch titled Revert "crypto: caam - get rid of tasklet" to the 4.9-stable tree which can be found at: http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=sum… The filename of the patch is: revert-crypto-caam-get-rid-of-tasklet.patch and it can be found in the queue-4.9 subdirectory. If you, or anyone else, feels it should not be added to the stable tree, please let <stable(a)vger.kernel.org> know about it. >From 2b163b5bce04546da72617bfb6c8bf07a45c4b17 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Horia=20Geant=C4=83?= <horia.geanta(a)nxp.com> Date: Wed, 9 Nov 2016 10:46:21 +0200 Subject: Revert "crypto: caam - get rid of tasklet" MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit From: Horia Geantă <horia.geanta(a)nxp.com> commit 2b163b5bce04546da72617bfb6c8bf07a45c4b17 upstream. This reverts commit 66d2e2028091a074aa1290d2eeda5ddb1a6c329c. Quoting from Russell's findings: https://www.mail-archive.com/linux-crypto@vger.kernel.org/msg21136.html [quote] Okay, I've re-tested, using a different way of measuring, because using openssl speed is impractical for off-loaded engines. I've decided to use this way to measure the performance: dd if=/dev/zero bs=1048576 count=128 | /usr/bin/time openssl dgst -md5 For the threaded IRQs case gives: 0.05user 2.74system 0:05.30elapsed 52%CPU (0avgtext+0avgdata 2400maxresident)k 0.06user 2.52system 0:05.18elapsed 49%CPU (0avgtext+0avgdata 2404maxresident)k 0.12user 2.60system 0:05.61elapsed 48%CPU (0avgtext+0avgdata 2460maxresident)k => 5.36s => 25.0MB/s and the tasklet case: 0.08user 2.53system 0:04.83elapsed 54%CPU (0avgtext+0avgdata 2468maxresident)k 0.09user 2.47system 0:05.16elapsed 49%CPU (0avgtext+0avgdata 2368maxresident)k 0.10user 2.51system 0:04.87elapsed 53%CPU (0avgtext+0avgdata 2460maxresident)k => 4.95 => 27.1MB/s which corresponds to an 8% slowdown for the threaded IRQ case. So, tasklets are indeed faster than threaded IRQs. [...] I think I've proven from the above that this patch needs to be reverted due to the performance regression, and that there _is_ most definitely a deterimental effect of switching from tasklets to threaded IRQs. [/quote] Signed-off-by: Horia Geantă <horia.geanta(a)nxp.com> Signed-off-by: Herbert Xu <herbert(a)gondor.apana.org.au> Signed-off-by: Greg Kroah-Hartman <gregkh(a)linuxfoundation.org> --- drivers/crypto/caam/intern.h | 1 + drivers/crypto/caam/jr.c | 25 ++++++++++++++++--------- 2 files changed, 17 insertions(+), 9 deletions(-) --- a/drivers/crypto/caam/intern.h +++ b/drivers/crypto/caam/intern.h @@ -41,6 +41,7 @@ struct caam_drv_private_jr { struct device *dev; int ridx; struct caam_job_ring __iomem *rregs; /* JobR's register space */ + struct tasklet_struct irqtask; int irq; /* One per queue */ /* Number of scatterlist crypt transforms active on the JobR */ --- a/drivers/crypto/caam/jr.c +++ b/drivers/crypto/caam/jr.c @@ -73,6 +73,8 @@ static int caam_jr_shutdown(struct devic ret = caam_reset_hw_jr(dev); + tasklet_kill(&jrp->irqtask); + /* Release interrupt */ free_irq(jrp->irq, dev); @@ -128,7 +130,7 @@ static irqreturn_t caam_jr_interrupt(int /* * Check the output ring for ready responses, kick - * the threaded irq if jobs done. + * tasklet if jobs done. */ irqstate = rd_reg32(&jrp->rregs->jrintstatus); if (!irqstate) @@ -150,13 +152,18 @@ static irqreturn_t caam_jr_interrupt(int /* Have valid interrupt at this point, just ACK and trigger */ wr_reg32(&jrp->rregs->jrintstatus, irqstate); - return IRQ_WAKE_THREAD; + preempt_disable(); + tasklet_schedule(&jrp->irqtask); + preempt_enable(); + + return IRQ_HANDLED; } -static irqreturn_t caam_jr_threadirq(int irq, void *st_dev) +/* Deferred service handler, run as interrupt-fired tasklet */ +static void caam_jr_dequeue(unsigned long devarg) { int hw_idx, sw_idx, i, head, tail; - struct device *dev = st_dev; + struct device *dev = (struct device *)devarg; struct caam_drv_private_jr *jrp = dev_get_drvdata(dev); void (*usercall)(struct device *dev, u32 *desc, u32 status, void *arg); u32 *userdesc, userstatus; @@ -230,8 +237,6 @@ static irqreturn_t caam_jr_threadirq(int /* reenable / unmask IRQs */ clrsetbits_32(&jrp->rregs->rconfig_lo, JRCFG_IMSK, 0); - - return IRQ_HANDLED; } /** @@ -389,10 +394,11 @@ static int caam_jr_init(struct device *d jrp = dev_get_drvdata(dev); + tasklet_init(&jrp->irqtask, caam_jr_dequeue, (unsigned long)dev); + /* Connect job ring interrupt handler. */ - error = request_threaded_irq(jrp->irq, caam_jr_interrupt, - caam_jr_threadirq, IRQF_SHARED, - dev_name(dev), dev); + error = request_irq(jrp->irq, caam_jr_interrupt, IRQF_SHARED, + dev_name(dev), dev); if (error) { dev_err(dev, "can't connect JobR %d interrupt (%d)\n", jrp->ridx, jrp->irq); @@ -454,6 +460,7 @@ out_free_inpring: out_free_irq: free_irq(jrp->irq, dev); out_kill_deq: + tasklet_kill(&jrp->irqtask); return error; } Patches currently in stable-queue which might be from horia.geanta(a)nxp.com are queue-4.9/revert-crypto-caam-get-rid-of-tasklet.patch

7 years, 1 month

1
0
0 0

[Linux-stable-mirror] [PATCH v3 05/15] drm/sun4i: Fix error path handling

by Maxime Ripard

The commit 4c7f16d14a33 ("drm/sun4i: Fix TCON clock and regmap initialization sequence") moved a bunch of logic around, but forgot to update the gotos after the introduction of the err_free_dotclock label. It means that if we fail later that the one introduced in that commit, we'll just to the old label which isn't free the clock we created. This will result in a breakage as soon as someone tries to do something with that clock, since its resources will have been long reclaimed. Cc: <stable(a)vger.kernel.org> Fixes: 4c7f16d14a33 ("drm/sun4i: Fix TCON clock and regmap initialization sequence") Signed-off-by: Maxime Ripard <maxime.ripard(a)free-electrons.com> --- drivers/gpu/drm/sun4i/sun4i_tcon.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/sun4i/sun4i_tcon.c b/drivers/gpu/drm/sun4i/sun4i_tcon.c index a1ed462c2430..ea056a3d2131 100644 --- a/drivers/gpu/drm/sun4i/sun4i_tcon.c +++ b/drivers/gpu/drm/sun4i/sun4i_tcon.c @@ -724,12 +724,12 @@ static int sun4i_tcon_bind(struct device *dev, struct device *master, if (IS_ERR(tcon->crtc)) { dev_err(dev, "Couldn't create our CRTC\n"); ret = PTR_ERR(tcon->crtc); - goto err_free_clocks; + goto err_free_dotclock; } ret = sun4i_rgb_init(drm, tcon); if (ret < 0) - goto err_free_clocks; + goto err_free_dotclock; if (tcon->quirks->needs_de_be_mux) { /* -- git-series 0.9.1

7 years, 1 month

2
1
0 0

[Linux-stable-mirror] [PATCH] usb: musb: da8xx: fix babble condition handling

by Bin Liu

When babble condition happens, the musb controller might automatically turns off VBUS. On DA8xx platform, the controller generates drvvbus interrupt for turning off VBUS along with the babble interrupt. In this case, we should handle the babble interrupt first and recover from the babble condition. This change ignores the drvvbus interrupt if babble interrupt is also generated at the same time, so the babble recovery routine works properly. Cc: stable(a)vger.kernel.org # v3.16+ Signed-off-by: Bin Liu <b-liu(a)ti.com> --- drivers/usb/musb/da8xx.c | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) diff --git a/drivers/usb/musb/da8xx.c b/drivers/usb/musb/da8xx.c index 0397606a211b..6c036de63272 100644 --- a/drivers/usb/musb/da8xx.c +++ b/drivers/usb/musb/da8xx.c @@ -284,7 +284,15 @@ static irqreturn_t da8xx_musb_interrupt(int irq, void *hci) musb->xceiv->otg->state = OTG_STATE_A_WAIT_VRISE; portstate(musb->port1_status |= USB_PORT_STAT_POWER); del_timer(&musb->dev_timer); - } else { + } else if (!(musb->int_usb & MUSB_INTR_BABBLE)) { + /* + * When babble condition happens, drvvbus interrupt + * is also generated. Ignore this drvvbus interrupt + * and let babble interrupt handler recovers the + * controller; otherwise, the host-mode flag is lost + * due to the MUSB_DEV_MODE() call below and babble + * recovery logic will not be called. + */ musb->is_active = 0; MUSB_DEV_MODE(musb); otg->default_a = 0; -- 1.9.1

7 years, 1 month

1
0
0 0

[Linux-stable-mirror] Patch "drm/fsl-dcu: enable IRQ before drm_atomic_helper_resume()" has been added to the 4.14-stable tree

by gregkh＠linuxfoundation.org

This is a note to let you know that I've just added the patch titled drm/fsl-dcu: enable IRQ before drm_atomic_helper_resume() to the 4.14-stable tree which can be found at: http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=sum… The filename of the patch is: drm-fsl-dcu-enable-irq-before-drm_atomic_helper_resume.patch and it can be found in the queue-4.14 subdirectory. If you, or anyone else, feels it should not be added to the stable tree, please let <stable(a)vger.kernel.org> know about it. >From 9fd99f4f3f5e13ce959900ae57d64b1bdb51d823 Mon Sep 17 00:00:00 2001 From: Stefan Agner <stefan(a)agner.ch> Date: Fri, 10 Nov 2017 10:15:28 +0100 Subject: drm/fsl-dcu: enable IRQ before drm_atomic_helper_resume() From: Stefan Agner <stefan(a)agner.ch> commit 9fd99f4f3f5e13ce959900ae57d64b1bdb51d823 upstream. The resume helpers wait for a vblank to occurre hence IRQ need to be enabled. This avoids a warning as follows during resume: WARNING: CPU: 0 PID: 314 at drivers/gpu/drm/drm_atomic_helper.c:1249 drm_atomic_helper_wait_for_vblanks.part.1+0x284/0x288 [CRTC:28:crtc-0] vblank wait timed out Signed-off-by: Stefan Agner <stefan(a)agner.ch> Signed-off-by: Greg Kroah-Hartman <gregkh(a)linuxfoundation.org> --- drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) --- a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c +++ b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c @@ -232,6 +232,7 @@ static int fsl_dcu_drm_pm_resume(struct if (fsl_dev->tcon) fsl_tcon_bypass_enable(fsl_dev->tcon); fsl_dcu_drm_init_planes(fsl_dev->drm); + enable_irq(fsl_dev->irq); drm_atomic_helper_resume(fsl_dev->drm, fsl_dev->state); console_lock(); @@ -239,7 +240,6 @@ static int fsl_dcu_drm_pm_resume(struct console_unlock(); drm_kms_helper_poll_enable(fsl_dev->drm); - enable_irq(fsl_dev->irq); return 0; } Patches currently in stable-queue which might be from stefan(a)agner.ch are queue-4.14/drm-fsl-dcu-avoid-disabling-pixel-clock-twice-on-suspend.patch queue-4.14/drm-fsl-dcu-enable-irq-before-drm_atomic_helper_resume.patch

7 years, 1 month

1
0
0 0

2025

2024

2023

2022

2021

2020

2019

2018

2017

Linux-stable-mirror December 2017