nvme pci: fix nvme_setup_irqs()
PATCH 1/1 nvme pci: fix nvme_setup_irqs()
When -ENOSPC is returned from pci_alloc_irq_vectors_affinity(), we still try to allocate multiple irq vectors again, so irq queues covers the admin queue actually. But we don't consider that, then number of the allocated irq vector may be same with sum of io_queues[HCTX_TYPE_DEFAULT] and io_queues[HCTX_TYPE_READ], this way is obviously wrong, and finally breaks nvme_pci_map_queues(), and warning from pci_irq_get_affinity() is triggered.
IRQ queues should cover admin queues, this patch makes this point explicitely in nvme_calc_io_queues().
We got severl boot failure internal report on aarch64, so please consider to fix it in v4.20.
--- drivers/nvme/host/pci.c | 21 +++++++++++++-------- 1 file changed, 13 insertions(+), 8 deletions(-)
drivers/nvme/host/pci.c +13 -8
@@ -2028,14 +2028,18 @@ static int nvme_setup_host_mem(struct nvme_dev *dev)return ret; } +/* irq_queues covers admin queue */ static void nvme_calc_io_queues (struct nvme_dev *dev,unsigned int irq_queues) {unsigned int this_w_queues = write_queues; +WARN_ON (!irq_queues); +/* -* Setup read/write queue split +* Setup read/write queue split, assign admin queue one independent +* irq vector if irq_queues is > 1. */ -if (irq_queues ==1 ) { +if (irq_queues <=2 ) { dev->io_queues[HCTX_TYPE_DEFAULT ] =1 ; dev->io_queues[HCTX_TYPE_READ ] =0 ;return ;
@@ -2043,21 +2047,21 @@ static void nvme_calc_io_queues(struct nvme_dev *dev, unsigned int irq_queues)/* * If 'write_queues' is set, ensure it leaves room for at least -* one read queue +* one read queue and one admin queue */ if (this_w_queues >= irq_queues) - this_w_queues = irq_queues -1 ; + this_w_queues = irq_queues -2 ;/* * If 'write_queues' is set to zero, reads and writes will share * a queue set. */ if (!this_w_queues) { - dev->io_queues[HCTX_TYPE_DEFAULT ] = irq_queues; + dev->io_queues[HCTX_TYPE_DEFAULT ] = irq_queues -1 ; dev->io_queues[HCTX_TYPE_READ ] =0 ; }else { dev->io_queues[HCTX_TYPE_DEFAULT ] = this_w_queues; - dev->io_queues[HCTX_TYPE_READ ] = irq_queues - this_w_queues; + dev->io_queues[HCTX_TYPE_READ ] = irq_queues - this_w_queues -1 ; } }
@@ -2082,7 +2086,7 @@ static int nvme_setup_irqs(struct nvme_dev *dev, unsigned int nr_io_queues) this_p_queues = nr_io_queues -1 ; irq_queues =1 ; }else { - irq_queues = nr_io_queues - this_p_queues; + irq_queues = nr_io_queues - this_p_queues +1 ; } dev->io_queues[HCTX_TYPE_POLL ] = this_p_queues;
@@ -2102,8 +2106,9 @@ static int nvme_setup_irqs(struct nvme_dev *dev, unsigned int nr_io_queues)* If we got a failure and we're down to asking for just * 1 + 1 queues, just ask for a single vector. We'll share * that between the single IO queue and the admin queue. +* Otherwise, we assign one independent vector to admin queue. */ -if (result >=0 && irq_queues >1 ) +if (irq_queues >1 ) irq_queues = irq_sets[0 ] + irq_sets[1 ] +1 ; result = pci_alloc_irq_vectors_affinity(pdev, irq_queues,
Signature · 1 line
2.9.5
PATCH 1/1 nvme pci: fix nvme_setup_irqs()
Quoted 96 lines · Ming Lei
When
-ENOSPCis returned frompci_alloc_irq_vectors_affinity(), we still try to allocate multiple irq vectors again, so irq queues covers the admin queue actually. But we don't consider that, then number of the allocated irq vector may be same with sum of io_queues[HCTX_TYPE_DEFAULT] and io_queues[HCTX_TYPE_READ], this way is obviously wrong, and finally breaksnvme_pci_map_queues(), and warning frompci_irq_get_affinity()is triggered.IRQ queues should cover admin queues, this patch makes this point explicitely in
nvme_calc_io_queues().We got severl boot failure internal report on aarch64, so please consider to fix it in v4.20.
Fixes 6451fe73fa0f ("nvme: fix irq vs io_queue calculations")Cc Keith Busch <keith.busch at intel.com>Cc Jens Axboe <axboe at fb.com>Signed-off-by Ming Lei <ming.lei at redhat.com>--- drivers/nvme/host/pci.c | 21 +++++++++++++-------- 1 file changed, 13 insertions(+), 8 deletions(-)drivers/nvme/host/pci.c +13
-8@@ -2028,14 +2028,18 @@ static int nvme_setup_host_mem(struct nvme_dev *dev)return ret; } +/* irq_queues covers admin queue */ static void nvme_calc_io_queues (struct nvme_dev *dev,unsigned int irq_queues) {unsigned int this_w_queues = write_queues; +WARN_ON (!irq_queues); +/* -* Setup read/write queue split +* Setup read/write queue split, assign admin queue one independent +* irq vector if irq_queues is > 1. */ -if (irq_queues ==1 ) { +if (irq_queues <=2 ) { dev->io_queues[HCTX_TYPE_DEFAULT ] =1 ; dev->io_queues[HCTX_TYPE_READ ] =0 ;return ;@@ -2043,21 +2047,21 @@ static void nvme_calc_io_queues(struct nvme_dev *dev, unsigned int irq_queues)/* * If 'write_queues' is set, ensure it leaves room for at least -* one read queue +* one read queue and one admin queue */ if (this_w_queues >= irq_queues) - this_w_queues = irq_queues -1 ; + this_w_queues = irq_queues -2 ;/* * If 'write_queues' is set to zero, reads and writes will share * a queue set. */ if (!this_w_queues) { - dev->io_queues[HCTX_TYPE_DEFAULT ] = irq_queues; + dev->io_queues[HCTX_TYPE_DEFAULT ] = irq_queues -1 ; dev->io_queues[HCTX_TYPE_READ ] =0 ; }else { dev->io_queues[HCTX_TYPE_DEFAULT ] = this_w_queues; - dev->io_queues[HCTX_TYPE_READ ] = irq_queues - this_w_queues; + dev->io_queues[HCTX_TYPE_READ ] = irq_queues - this_w_queues -1 ; } }@@ -2082,7 +2086,7 @@ static int nvme_setup_irqs(struct nvme_dev *dev, unsigned int nr_io_queues) this_p_queues = nr_io_queues -1 ; irq_queues =1 ; }else { - irq_queues = nr_io_queues - this_p_queues; + irq_queues = nr_io_queues - this_p_queues +1 ; } dev->io_queues[HCTX_TYPE_POLL ] = this_p_queues;@@ -2102,8 +2106,9 @@ static int nvme_setup_irqs(struct nvme_dev *dev, unsigned int nr_io_queues)* If we got a failure and we're down to asking for just * 1 + 1 queues, just ask for a single vector. We'll share * that between the single IO queue and the admin queue. +* Otherwise, we assign one independent vector to admin queue. */ -if (result >=0 && irq_queues >1 ) +if (irq_queues >1 ) irq_queues = irq_sets[0 ] + irq_sets[1 ] +1 ; result = pci_alloc_irq_vectors_affinity(pdev, irq_queues,Signature · 1 line
2.9.5
Ping...
Thanks,
Ming
PATCH 1/1 nvme pci: fix nvme_setup_irqs()
In reply to Ming LeiQuoted 8 lines · Ming Lei
When
-ENOSPCis returned frompci_alloc_irq_vectors_affinity(), we still try to allocate multiple irq vectors again, so irq queues covers the admin queue actually. But we don't consider that, then number of the allocated irq vector may be same with sum of io_queues[HCTX_TYPE_DEFAULT] and io_queues[HCTX_TYPE_READ], this way is obviously wrong, and finally breaksnvme_pci_map_queues(), and warning frompci_irq_get_affinity()is triggered.
Quoted 5 lines
IRQ queues should cover admin queues, this patch makes this point explicitely in
nvme_calc_io_queues().We got severl boot failure internal report on aarch64, so please consider to fix it in v4.20.
I see what you saying with the inconsistent meaning for irq_queues, though 4.20 should be fine.
I hope we can make the irq sets easier to use in the future, but your patch looks correct for the current interface.
Quoted 4 lines
Fixes 6451fe73fa0f ("nvme: fix irq vs io_queue calculations")Cc Keith Busch <keith.busch at intel.com>Cc Jens Axboe <axboe at fb.com>Signed-off-by Ming Lei <ming.lei at redhat.com>
PATCH 1/1 nvme pci: fix nvme_setup_irqs()
Quoted 12 lines · Keith Busch
Quoted 11 lines · Ming Lei
When
-ENOSPCis returned frompci_alloc_irq_vectors_affinity(), we still try to allocate multiple irq vectors again, so irq queues covers the admin queue actually. But we don't consider that, then number of the allocated irq vector may be same with sum of io_queues[HCTX_TYPE_DEFAULT] and io_queues[HCTX_TYPE_READ], this way is obviously wrong, and finally breaksnvme_pci_map_queues(), and warning frompci_irq_get_affinity()is triggered.IRQ queues should cover admin queues, this patch makes this point explicitely in
nvme_calc_io_queues().
Quoted 7 lines
Quoted 2 lines
We got severl boot failure internal report on aarch64, so please consider to fix it in v4.20.
I see what you saying with the inconsistent meaning for
irq_queues, though 4.20 should be fine.I hope we can make the irq sets easier to use in the future, but your
Yeah, I agree the current API isn't easy to use for irq sets, and we need to improve this.
Quoted 3 lines
patch looks correct for the current interface.
Reviewed-by Keith Busch <keith.busch at intel.com>
Thanks!
Christoph, what do you think of this patch?
Thanks,
Ming
PATCH 1/1 nvme pci: fix nvme_setup_irqs()
In reply to Ming LeiThanks,
applied to nvme-5.0.