May 2018 - Linux-stable-mirror

by Fabrizio Castro

Hello Greg, Thank you for your feedback. > Subject: Re: [PATCH] dmaengine: ensure dmaengine helpers check valid callback > > On Mon, May 21, 2018 at 05:56:55PM +0100, Fabrizio Castro wrote: > > From: Vinod Koul <vinod.koul(a)intel.com> > > > > commit 757d12e5849be549076901b0d33c60d5f360269c upstream. > > > > dmaengine has various device callbacks and exposes helper > > functions to invoke these. These helpers should check if channel, > > device and callback is valid or not before invoking them. > > > > Reported-by: Jon Hunter <jonathanh(a)nvidia.com> > > Signed-off-by: Vinod Koul <vinod.koul(a)intel.com> > > [fabrizio: cherry-pick to 4.4] > > Signed-off-by: Fabrizio Castro <fabrizio.castro(a)bp.renesas.com> > > Signed-off-by: Jianming Qiao <jianming.qiao(a)bp.renesas.com> > > --- > > Hello Greg, > > > > while backporting commit 757d12e5849be549076901b0d33c60d5f360269c > > to the CIP kernel Ben recommended to send the same patch to you > > for 4.4 stable. > > I hope the format of the commit is the one you expect (reference to > > the upstream commit, version to cherry-pick the patch to, and > > Signed-off-by tags). > > Format is fine, but why is this needed in the 4.4.y kernel tree? We work with the CIP kernel (v4.4), and from time to time we come across bug fixes we feel like stable could benefit from. Also, since Ben merges the CIP branch with stable, the fixes we make to stable will appear in the CIP kernel too at some point. If you feel like the patch is not worth considering for stable, it can still be applied to the CIP kernel if required. Ben has already taken this patch (for v4.4.126-cip22) therefore it's not on our critical path, but it would be nice to have in the stable kernel too. > What bug does it solve? Without this patch we managed to get the kernel to try and dereference a NULL pointer while playing around with ttys (basically by having the wrong tty with /etc/securetty). Are you happy to take this patch? Thanks, Fab > > thanks, > > greg k-h Renesas Electronics Europe Ltd, Dukes Meadow, Millboard Road, Bourne End, Buckinghamshire, SL8 5FH, UK. Registered in England & Wales under Registered No. 04586709.

6 years, 7 months

2
2
0 0

[PATCH] usb: typec: wcove: Remove dependency on HW FSM

by Heikki Krogerus

The USB Type-C PHY in Intel WhiskeyCove PMIC has build-in USB Type-C state machine which we were relying on to configure the CC lines correctly. This patch removes that dependency and configures the CC line according to commands from the port manager (tcpm.c) in wcove_set_cc(). This fixes an issue where USB devices attached to the USB Type-C port do not get enumerated. When acting as source/host, the HW FSM sometimes fails to configure the PHY correctly. Fixes: 3c4fb9f16921 ("usb: typec: wcove: start using tcpm for USB PD support") Cc: stable(a)vger.kernel.org Signed-off-by: Heikki Krogerus <heikki.krogerus(a)linux.intel.com> --- drivers/usb/typec/typec_wcove.c | 30 ++++++++++++++++++++++++++++-- 1 file changed, 28 insertions(+), 2 deletions(-) diff --git a/drivers/usb/typec/typec_wcove.c b/drivers/usb/typec/typec_wcove.c index 39cff11ec7a2..423208e19383 100644 --- a/drivers/usb/typec/typec_wcove.c +++ b/drivers/usb/typec/typec_wcove.c @@ -202,6 +202,10 @@ static int wcove_init(struct tcpc_dev *tcpc) struct wcove_typec *wcove = tcpc_to_wcove(tcpc); int ret; + ret = regmap_write(wcove->regmap, USBC_CONTROL1, 0); + if (ret) + return ret; + /* Unmask everything */ ret = regmap_write(wcove->regmap, USBC_IRQMASK1, 0); if (ret) @@ -285,8 +289,30 @@ static int wcove_get_cc(struct tcpc_dev *tcpc, enum typec_cc_status *cc1, static int wcove_set_cc(struct tcpc_dev *tcpc, enum typec_cc_status cc) { - /* XXX: Relying on the HW FSM to configure things correctly for now */ - return 0; + struct wcove_typec *wcove = tcpc_to_wcove(tcpc); + unsigned int ctrl; + + switch (cc) { + case TYPEC_CC_RD: + ctrl = USBC_CONTROL1_MODE_SNK; + break; + case TYPEC_CC_RP_DEF: + ctrl = USBC_CONTROL1_CURSRC_UA_80 | USBC_CONTROL1_MODE_SRC; + break; + case TYPEC_CC_RP_1_5: + ctrl = USBC_CONTROL1_CURSRC_UA_180 | USBC_CONTROL1_MODE_SRC; + break; + case TYPEC_CC_RP_3_0: + ctrl = USBC_CONTROL1_CURSRC_UA_330 | USBC_CONTROL1_MODE_SRC; + break; + case TYPEC_CC_OPEN: + ctrl = 0; + break; + default: + return -EINVAL; + } + + return regmap_write(wcove->regmap, USBC_CONTROL1, ctrl); } static int wcove_set_polarity(struct tcpc_dev *tcpc, enum typec_cc_polarity pol) -- 2.17.0

6 years, 7 months

1
0
0 0

[GIT PULL 2/2] intel_th: Use correct device when freeing buffers

by Alexander Shishkin

Commit d5c435df4a890 ("intel_th: msu: Use the real device in case of IOMMU domain allocation") changes dma buffer allocation to use the actual underlying device, but forgets to change the deallocation path, which leads to (if you've got CAP_SYS_RAWIO): > # echo 0,0 > /sys/bus/intel_th/devices/0-msc0/nr_pages > ------------[ cut here ]------------ > kernel BUG at ../linux/drivers/iommu/intel-iommu.c:3670! > CPU: 3 PID: 231 Comm: sh Not tainted 4.17.0-rc1+ #2729 > RIP: 0010:intel_unmap+0x11e/0x130 ... > Call Trace: > intel_free_coherent+0x3e/0x60 > msc_buffer_win_free+0x100/0x160 [intel_th_msu] This patch fixes the buffer deallocation code to use the correct device. Signed-off-by: Alexander Shishkin <alexander.shishkin(a)linux.intel.com> Fixes: d5c435df4a890 ("intel_th: msu: Use the real device in case of IOMMU domain allocation") Reported-by: Baofeng Tian <baofeng.tian(a)intel.com> CC: stable(a)vger.kernel.org # v4.14+ --- drivers/hwtracing/intel_th/msu.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/drivers/hwtracing/intel_th/msu.c b/drivers/hwtracing/intel_th/msu.c index ede388309376..634f58042c77 100644 --- a/drivers/hwtracing/intel_th/msu.c +++ b/drivers/hwtracing/intel_th/msu.c @@ -733,8 +733,8 @@ static int msc_buffer_win_alloc(struct msc *msc, unsigned int nr_blocks) /* Reset the page to write-back before releasing */ set_memory_wb((unsigned long)win->block[i].bdesc, 1); #endif - dma_free_coherent(msc_dev(msc), size, win->block[i].bdesc, - win->block[i].addr); + dma_free_coherent(msc_dev(msc)->parent->parent, size, + win->block[i].bdesc, win->block[i].addr); } kfree(win); @@ -769,7 +769,7 @@ static void msc_buffer_win_free(struct msc *msc, struct msc_window *win) /* Reset the page to write-back before releasing */ set_memory_wb((unsigned long)win->block[i].bdesc, 1); #endif - dma_free_coherent(msc_dev(win->msc), PAGE_SIZE, + dma_free_coherent(msc_dev(win->msc)->parent->parent, PAGE_SIZE, win->block[i].bdesc, win->block[i].addr); } -- 2.17.0

6 years, 7 months

1
0
0 0

[GIT PULL 1/2] stm class: Use vmalloc for the master map

by Alexander Shishkin

Fengguang is running into a warning from the buddy allocator: > swapper/0: page allocation failure: order:9, mode:0x14040c0(GFP_KERNEL|__GFP_COMP), nodemask=(null) > CPU: 1 PID: 1 Comm: swapper/0 Not tainted 4.17.0-rc1 #262 > Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.10.2-1 04/01/2014 > Call Trace: ... > __kmalloc+0x14b/0x180: ____cache_alloc at mm/slab.c:3127 > stm_register_device+0xf3/0x5c0: stm_register_device at drivers/hwtracing/stm/core.c:695 ... Which is basically a result of the stm class trying to allocate ~512kB for the dummy_stm with its default parameters. There's no reason, however, for it not to be vmalloc()ed instead, which is what this patch does. Reported-by: Fengguang Wu <fengguang.wu(a)intel.com> Signed-off-by: Alexander Shishkin <alexander.shishkin(a)linux.intel.com> CC: stable(a)vger.kernel.org # v4.4+ --- drivers/hwtracing/stm/core.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/drivers/hwtracing/stm/core.c b/drivers/hwtracing/stm/core.c index 05386b76465e..657badb479a5 100644 --- a/drivers/hwtracing/stm/core.c +++ b/drivers/hwtracing/stm/core.c @@ -674,7 +674,7 @@ static void stm_device_release(struct device *dev) { struct stm_device *stm = to_stm_device(dev); - kfree(stm); + vfree(stm); } int stm_register_device(struct device *parent, struct stm_data *stm_data, @@ -691,7 +691,7 @@ int stm_register_device(struct device *parent, struct stm_data *stm_data, return -EINVAL; nmasters = stm_data->sw_end - stm_data->sw_start + 1; - stm = kzalloc(sizeof(*stm) + nmasters * sizeof(void *), GFP_KERNEL); + stm = vzalloc(sizeof(*stm) + nmasters * sizeof(void *)); if (!stm) return -ENOMEM; @@ -744,7 +744,7 @@ int stm_register_device(struct device *parent, struct stm_data *stm_data, /* matches device_initialize() above */ put_device(&stm->dev); err_free: - kfree(stm); + vfree(stm); return err; } -- 2.17.0

6 years, 7 months

1
0
0 0

[PATCH] pinctrl: armada-37xx: Fix spurious irq management

by Gregory CLEMENT

From: Terry Zhou <bjzhou(a)marvell.com> Until now, if we found spurious irq in irq_handler, we only updated the status in register but not the status in the code. Due to this the system will got stuck dues to the infinite loop [gregory.clement(a)bootlin.com: update comment and add fix and stable tags] Fixes: 30ac0d3b0702 ("pinctrl: armada-37xx: Add edge both type gpio irq support") Cc: <stable(a)vger.kernel.org> Signed-off-by: Terry Zhou <bjzhou(a)marvell.com> Reviewed-by: Gregory CLEMENT <gregory.clement(a)bootlin.com> Signed-off-by: Gregory CLEMENT <gregory.clement(a)bootlin.com> --- drivers/pinctrl/mvebu/pinctrl-armada-37xx.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/drivers/pinctrl/mvebu/pinctrl-armada-37xx.c b/drivers/pinctrl/mvebu/pinctrl-armada-37xx.c index 5b63248c8209..7bef929bd7fe 100644 --- a/drivers/pinctrl/mvebu/pinctrl-armada-37xx.c +++ b/drivers/pinctrl/mvebu/pinctrl-armada-37xx.c @@ -679,12 +679,13 @@ static void armada_37xx_irq_handler(struct irq_desc *desc) writel(1 << hwirq, info->base + IRQ_STATUS + 4 * i); - continue; + goto update_status; } } generic_handle_irq(virq); +update_status: /* Update status in case a new IRQ appears */ spin_lock_irqsave(&info->irq_lock, flags); status = readl_relaxed(info->base + -- 2.17.0

6 years, 7 months

2
1
0 0

[PATCH] gpio: rcar: Add Runtime PM handling for interrupts

by Fabrizio Castro

From: Geert Uytterhoeven <geert+renesas(a)glider.be> commit b26a719bdba9aa926ceaadecc66e07623d2b8a53 upstream. The R-Car GPIO driver handles Runtime PM for requested GPIOs only. When using a GPIO purely as an interrupt source, no Runtime PM handling is done, and the GPIO module's clock may not be enabled. To fix this: - Add .irq_request_resources() and .irq_release_resources() callbacks to handle Runtime PM when an interrupt is requested, - Add irq_bus_lock() and sync_unlock() callbacks to handle Runtime PM when e.g. disabling/enabling an interrupt, or configuring the interrupt type. Fixes: d5c3d84657db57bd "net: phy: Avoid polling PHY with PHY_IGNORE_INTERRUPTS" Signed-off-by: Geert Uytterhoeven <geert+renesas(a)glider.be> Signed-off-by: Linus Walleij <linus.walleij(a)linaro.org> [fabrizio: cherry-pick to v4.4.y. Use container_of instead of gpiochip_get_data.] Signed-off-by: Fabrizio Castro <fabrizio.castro(a)bp.renesas.com> Reviewed-by: Biju Das <biju.das(a)bp.renesas.com> --- Hi Greg, on R-Car we have found that if a GPIO is used purely as an interrupt source, the corresponding clock is not enabled, therefore the interrupt doesn't work. On Koelsch, the HDMI trasmitter node in the DT uses GPIO 3 29 for the interrupt line, and since gpio3 is used only for this, its clock is OFF, preventing the interrupt from working. This patch fixes this problem. Would you please consider this patch for 4.4.y? Thanks, Fab drivers/gpio/gpio-rcar.c | 46 ++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 46 insertions(+) diff --git a/drivers/gpio/gpio-rcar.c b/drivers/gpio/gpio-rcar.c index 2a81224..9ba4aaa 100644 --- a/drivers/gpio/gpio-rcar.c +++ b/drivers/gpio/gpio-rcar.c @@ -200,6 +200,48 @@ static int gpio_rcar_irq_set_wake(struct irq_data *d, unsigned int on) return 0; } +static void gpio_rcar_irq_bus_lock(struct irq_data *d) +{ + struct gpio_chip *gc = irq_data_get_irq_chip_data(d); + struct gpio_rcar_priv *p = container_of(gc, struct gpio_rcar_priv, + gpio_chip); + + pm_runtime_get_sync(&p->pdev->dev); +} + +static void gpio_rcar_irq_bus_sync_unlock(struct irq_data *d) +{ + struct gpio_chip *gc = irq_data_get_irq_chip_data(d); + struct gpio_rcar_priv *p = container_of(gc, struct gpio_rcar_priv, + gpio_chip); + + pm_runtime_put(&p->pdev->dev); +} + + +static int gpio_rcar_irq_request_resources(struct irq_data *d) +{ + struct gpio_chip *gc = irq_data_get_irq_chip_data(d); + struct gpio_rcar_priv *p = container_of(gc, struct gpio_rcar_priv, + gpio_chip); + int error; + + error = pm_runtime_get_sync(&p->pdev->dev); + if (error < 0) + return error; + + return 0; +} + +static void gpio_rcar_irq_release_resources(struct irq_data *d) +{ + struct gpio_chip *gc = irq_data_get_irq_chip_data(d); + struct gpio_rcar_priv *p = container_of(gc, struct gpio_rcar_priv, + gpio_chip); + + pm_runtime_put(&p->pdev->dev); +} + static irqreturn_t gpio_rcar_irq_handler(int irq, void *dev_id) { struct gpio_rcar_priv *p = dev_id; @@ -460,6 +502,10 @@ static int gpio_rcar_probe(struct platform_device *pdev) irq_chip->irq_unmask = gpio_rcar_irq_enable; irq_chip->irq_set_type = gpio_rcar_irq_set_type; irq_chip->irq_set_wake = gpio_rcar_irq_set_wake; + irq_chip->irq_bus_lock = gpio_rcar_irq_bus_lock; + irq_chip->irq_bus_sync_unlock = gpio_rcar_irq_bus_sync_unlock; + irq_chip->irq_request_resources = gpio_rcar_irq_request_resources; + irq_chip->irq_release_resources = gpio_rcar_irq_release_resources; irq_chip->flags = IRQCHIP_SET_TYPE_MASKED | IRQCHIP_MASK_ON_SUSPEND; ret = gpiochip_add(gpio_chip); -- 2.7.4

6 years, 7 months

2
1
0 0

Please apply 318aaf34f117 ("scsi: libsas: defer ata device eh...") to stable releases

by Guenter Roeck

Hi Greg, commit 318aaf34f1179b39f ("scsi: libsas: defer ata device eh commands to libata") fixes CVE-2018-10021. Its severity is disputed, yet it is a real bug. Please consider applying it to stable releases. Thanks, Guenter

6 years, 7 months

2
1
0 0

[PATCH 0/5] mm: rework hmm to use devm_memremap_pages

by Dan Williams

Hi Andrew, please consider this series for 4.18. For maintainability, as ZONE_DEVICE continues to attract new users, it is useful to keep all users consolidated on devm_memremap_pages() as the interface for create "device pages". The devm_memremap_pages() implementation was recently reworked to make it more generic for arbitrary users, like the proposed peer-to-peer PCI-E enabling. HMM pre-dated this rework and opted to duplicate devm_memremap_pages() as hmm_devmem_pages_create(). Rework HMM to be a consumer of devm_memremap_pages() directly and fix up the licensing on the exports given the deep dependencies on the mm. Patches based on v4.17-rc6 where there are no upstream consumers of the HMM functionality. --- Dan Williams (5): mm, devm_memremap_pages: mark devm_memremap_pages() EXPORT_SYMBOL_GPL mm, devm_memremap_pages: handle errors allocating final devres action mm, hmm: use devm semantics for hmm_devmem_{add,remove} mm, hmm: replace hmm_devmem_pages_create() with devm_memremap_pages() mm, hmm: mark hmm_devmem_{add,add_resource} EXPORT_SYMBOL_GPL Documentation/vm/hmm.txt | 1 include/linux/hmm.h | 4 - include/linux/memremap.h | 1 kernel/memremap.c | 39 +++++- mm/hmm.c | 297 +++++++--------------------------------------- 5 files changed, 77 insertions(+), 265 deletions(-)

6 years, 7 months

5
12
0 0

[PATCH V3] blk-mq: avoid to starve tag allocation after allocation process migrates

by Ming Lei

When the allocation process is scheduled back and the mapped hw queue is changed, fake one extra wake up on previous queue for compensating wake up miss, so other allocations on the previous queue won't be starved. This patch fixes one request allocation hang issue, which can be triggered easily in case of very low nr_request. Cc: <stable(a)vger.kernel.org> Cc: Omar Sandoval <osandov(a)fb.com> Signed-off-by: Ming Lei <ming.lei(a)redhat.com> --- V3: - fix comments as suggested by Jens - remove the wrapper as suggested by Omar V2: - fix build failure block/blk-mq-tag.c | 12 ++++++++++++ include/linux/sbitmap.h | 7 +++++++ lib/sbitmap.c | 22 ++++++++++++---------- 3 files changed, 31 insertions(+), 10 deletions(-) diff --git a/block/blk-mq-tag.c b/block/blk-mq-tag.c index 336dde07b230..a4e58fc28a06 100644 --- a/block/blk-mq-tag.c +++ b/block/blk-mq-tag.c @@ -134,6 +134,8 @@ unsigned int blk_mq_get_tag(struct blk_mq_alloc_data *data) ws = bt_wait_ptr(bt, data->hctx); drop_ctx = data->ctx == NULL; do { + struct sbitmap_queue *bt_prev; + /* * We're out of tags on this hardware queue, kick any * pending IO submits before going to sleep waiting for @@ -159,6 +161,7 @@ unsigned int blk_mq_get_tag(struct blk_mq_alloc_data *data) if (data->ctx) blk_mq_put_ctx(data->ctx); + bt_prev = bt; io_schedule(); data->ctx = blk_mq_get_ctx(data->q); @@ -170,6 +173,15 @@ unsigned int blk_mq_get_tag(struct blk_mq_alloc_data *data) bt = &tags->bitmap_tags; finish_wait(&ws->wait, &wait); + + /* + * If destination hw queue is changed, fake wake up on + * previous queue for compensating the wake up miss, so + * other allocations on previous queue won't be starved. + */ + if (bt != bt_prev) + sbitmap_queue_wake_up(bt_prev); + ws = bt_wait_ptr(bt, data->hctx); } while (1); diff --git a/include/linux/sbitmap.h b/include/linux/sbitmap.h index 841585f6e5f2..bba9d80191b7 100644 --- a/include/linux/sbitmap.h +++ b/include/linux/sbitmap.h @@ -484,6 +484,13 @@ static inline struct sbq_wait_state *sbq_wait_ptr(struct sbitmap_queue *sbq, void sbitmap_queue_wake_all(struct sbitmap_queue *sbq); /** + * sbitmap_queue_wake_up() - Wake up some of waiters in one waitqueue + * on a &struct sbitmap_queue. + * @sbq: Bitmap queue to wake up. + */ +void sbitmap_queue_wake_up(struct sbitmap_queue *sbq); + +/** * sbitmap_queue_show() - Dump &struct sbitmap_queue information to a &struct * seq_file. * @sbq: Bitmap queue to show. diff --git a/lib/sbitmap.c b/lib/sbitmap.c index e6a9c06ec70c..14e027a33ffa 100644 --- a/lib/sbitmap.c +++ b/lib/sbitmap.c @@ -335,8 +335,9 @@ void sbitmap_queue_resize(struct sbitmap_queue *sbq, unsigned int depth) if (sbq->wake_batch != wake_batch) { WRITE_ONCE(sbq->wake_batch, wake_batch); /* - * Pairs with the memory barrier in sbq_wake_up() to ensure that - * the batch size is updated before the wait counts. + * Pairs with the memory barrier in sbitmap_queue_wake_up() + * to ensure that the batch size is updated before the wait + * counts. */ smp_mb__before_atomic(); for (i = 0; i < SBQ_WAIT_QUEUES; i++) @@ -425,7 +426,7 @@ static struct sbq_wait_state *sbq_wake_ptr(struct sbitmap_queue *sbq) return NULL; } -static void sbq_wake_up(struct sbitmap_queue *sbq) +void sbitmap_queue_wake_up(struct sbitmap_queue *sbq) { struct sbq_wait_state *ws; unsigned int wake_batch; @@ -454,23 +455,24 @@ static void sbq_wake_up(struct sbitmap_queue *sbq) */ smp_mb__before_atomic(); /* - * If there are concurrent callers to sbq_wake_up(), the last - * one to decrement the wait count below zero will bump it back - * up. If there is a concurrent resize, the count reset will - * either cause the cmpxchg to fail or overwrite after the - * cmpxchg. + * If there are concurrent callers to sbitmap_queue_wake_up(), + * the last one to decrement the wait count below zero will + * bump it back up. If there is a concurrent resize, the count + * reset will either cause the cmpxchg to fail or overwrite + * after the cmpxchg. */ atomic_cmpxchg(&ws->wait_cnt, wait_cnt, wait_cnt + wake_batch); sbq_index_atomic_inc(&sbq->wake_index); wake_up_nr(&ws->wait, wake_batch); } } +EXPORT_SYMBOL_GPL(sbitmap_queue_wake_up); void sbitmap_queue_clear(struct sbitmap_queue *sbq, unsigned int nr, unsigned int cpu) { sbitmap_clear_bit_unlock(&sbq->sb, nr); - sbq_wake_up(sbq); + sbitmap_queue_wake_up(sbq); if (likely(!sbq->round_robin && nr < sbq->sb.depth)) *per_cpu_ptr(sbq->alloc_hint, cpu) = nr; } @@ -482,7 +484,7 @@ void sbitmap_queue_wake_all(struct sbitmap_queue *sbq) /* * Pairs with the memory barrier in set_current_state() like in - * sbq_wake_up(). + * sbitmap_queue_wake_up(). */ smp_mb(); wake_index = atomic_read(&sbq->wake_index); -- 2.9.5

6 years, 7 months

2
1
0 0

[PATCH V2] blk-mq: avoid to starve tag allocation after allocation process migrates

by Ming Lei

When the allocation process is scheduled back and the mapped hw queue is changed, do one extra wake up on orignal queue for compensating wake up miss, so other allocations on the orignal queue won't be starved. This patch fixes one request allocation hang issue, which can be triggered easily in case of very low nr_request. Cc: <stable(a)vger.kernel.org> Cc: Omar Sandoval <osandov(a)fb.com> Signed-off-by: Ming Lei <ming.lei(a)redhat.com> --- V2: fix build failure block/blk-mq-tag.c | 13 +++++++++++++ include/linux/sbitmap.h | 7 +++++++ lib/sbitmap.c | 6 ++++++ 3 files changed, 26 insertions(+) diff --git a/block/blk-mq-tag.c b/block/blk-mq-tag.c index 336dde07b230..77607f89d205 100644 --- a/block/blk-mq-tag.c +++ b/block/blk-mq-tag.c @@ -134,6 +134,8 @@ unsigned int blk_mq_get_tag(struct blk_mq_alloc_data *data) ws = bt_wait_ptr(bt, data->hctx); drop_ctx = data->ctx == NULL; do { + struct sbitmap_queue *bt_orig; + /* * We're out of tags on this hardware queue, kick any * pending IO submits before going to sleep waiting for @@ -159,6 +161,7 @@ unsigned int blk_mq_get_tag(struct blk_mq_alloc_data *data) if (data->ctx) blk_mq_put_ctx(data->ctx); + bt_orig = bt; io_schedule(); data->ctx = blk_mq_get_ctx(data->q); @@ -170,6 +173,16 @@ unsigned int blk_mq_get_tag(struct blk_mq_alloc_data *data) bt = &tags->bitmap_tags; finish_wait(&ws->wait, &wait); + + /* + * If destination hw queue is changed, wake up original + * queue one extra time for compensating the wake up + * miss, so other allocations on original queue won't + * be starved. + */ + if (bt != bt_orig) + sbitmap_queue_wake_up(bt_orig); + ws = bt_wait_ptr(bt, data->hctx); } while (1); diff --git a/include/linux/sbitmap.h b/include/linux/sbitmap.h index 841585f6e5f2..b23f50355281 100644 --- a/include/linux/sbitmap.h +++ b/include/linux/sbitmap.h @@ -484,6 +484,13 @@ static inline struct sbq_wait_state *sbq_wait_ptr(struct sbitmap_queue *sbq, void sbitmap_queue_wake_all(struct sbitmap_queue *sbq); /** + * sbitmap_wake_up() - Do a regular wake up compensation if the queue + * allocated from is changed after scheduling back. + * @sbq: Bitmap queue to wake up. + */ +void sbitmap_queue_wake_up(struct sbitmap_queue *sbq); + +/** * sbitmap_queue_show() - Dump &struct sbitmap_queue information to a &struct * seq_file. * @sbq: Bitmap queue to show. diff --git a/lib/sbitmap.c b/lib/sbitmap.c index e6a9c06ec70c..c6ae4206bcb1 100644 --- a/lib/sbitmap.c +++ b/lib/sbitmap.c @@ -466,6 +466,12 @@ static void sbq_wake_up(struct sbitmap_queue *sbq) } } +void sbitmap_queue_wake_up(struct sbitmap_queue *sbq) +{ + sbq_wake_up(sbq); +} +EXPORT_SYMBOL_GPL(sbitmap_queue_wake_up); + void sbitmap_queue_clear(struct sbitmap_queue *sbq, unsigned int nr, unsigned int cpu) { -- 2.9.5

6 years, 7 months

4
5
0 0

2024

2023

2022

2021

2020

2019

2018

2017

Linux-stable-mirror May 2018