Re: [PATCH V3 1/5] genirq/affinity: don't mark 'affd' as const

2019-02-19 Thread Keith Busch
On Mon, Feb 18, 2019 at 04:42:27PM -0800, 陈华才 wrote: > I've tested, this patch can fix the nvme problem, but it can't be applied > to 4.19 because of different context. And, I still think my original solution > (genirq/affinity: Assign default affinity to pre/post vectors) is correct. > There may b

Re: [PATCH V3 1/5] genirq/affinity: don't mark 'affd' as const

2019-02-18 Thread Thomas Gleixner
On Tue, 19 Feb 2019, 陈华才 wrote: > > I've tested, this patch can fix the nvme problem, but it can't be applied > to 4.19 because of different context. And, I still think my original solution > (genirq/affinity: Assign default affinity to pre/post vectors) is correct. > There may be similar problems

Re: [PATCH V3 1/5] genirq/affinity: don't mark 'affd' as const

2019-02-18 Thread 陈华才
;Jens Axboe"; "Sagi Grimberg"; "linux-pci"; "LKML"; "linux-nvme"; "Ming Lei"; "linux-block"; "Christoph Hellwig"; Subject: Re: [PATCH V3 1/5] genirq/affinity: don't mark 'affd' as const On Thu, 14 Feb 201

Re: [PATCH V3 1/5] genirq/affinity: don't mark 'affd' as const

2019-02-14 Thread Thomas Gleixner
On Thu, 14 Feb 2019, 陈华才 wrote: Please do not top post > I'll test next week, but 4.19 has the same problem, how to fix that for 4.19? By applying the very same patch perhaps? Thanks, tglx

Re: [PATCH V3 1/5] genirq/affinity: don't mark 'affd' as const

2019-02-14 Thread 陈华才
uot;Sagi Grimberg"; "linux-pci"; "LKML"; "linux-nvme"; "Ming Lei"; "linux-block"; "Christoph Hellwig"; "Huacai Chen"; Subject: Re: [PATCH V3 1/5] genirq/affinity: don't mark 'affd' as const On Wed, 13

Re: [PATCH V3 1/5] genirq/affinity: don't mark 'affd' as const

2019-02-14 Thread Thomas Gleixner
On Wed, 13 Feb 2019, Keith Busch wrote: Cc+ Huacai Chen > On Wed, Feb 13, 2019 at 10:41:55PM +0100, Thomas Gleixner wrote: > > Btw, while I have your attention. There popped up an issue recently related > > to that affinity logic. > > > > The current implementation fails when: > > > > /

Re: [PATCH V3 1/5] genirq/affinity: don't mark 'affd' as const

2019-02-13 Thread Keith Busch
On Wed, Feb 13, 2019 at 10:41:55PM +0100, Thomas Gleixner wrote: > Btw, while I have your attention. There popped up an issue recently related > to that affinity logic. > > The current implementation fails when: > > /* > * If there aren't any vectors left after applying the pre/p

Re: [PATCH V3 1/5] genirq/affinity: don't mark 'affd' as const

2019-02-13 Thread Thomas Gleixner
On Wed, 13 Feb 2019, Keith Busch wrote: > On Wed, Feb 13, 2019 at 09:56:36PM +0100, Thomas Gleixner wrote: > > On Wed, 13 Feb 2019, Bjorn Helgaas wrote: > > > On Wed, Feb 13, 2019 at 06:50:37PM +0800, Ming Lei wrote: > > > > We have to ask driver to re-caculate set vectors after the whole IRQ > > >

Re: [PATCH V3 1/5] genirq/affinity: don't mark 'affd' as const

2019-02-13 Thread Keith Busch
On Wed, Feb 13, 2019 at 09:56:36PM +0100, Thomas Gleixner wrote: > On Wed, 13 Feb 2019, Bjorn Helgaas wrote: > > On Wed, Feb 13, 2019 at 06:50:37PM +0800, Ming Lei wrote: > > > We have to ask driver to re-caculate set vectors after the whole IRQ > > > vectors are allocated later, and the result nee

Re: [PATCH V3 1/5] genirq/affinity: don't mark 'affd' as const

2019-02-13 Thread Thomas Gleixner
On Wed, 13 Feb 2019, Bjorn Helgaas wrote: > On Wed, Feb 13, 2019 at 06:50:37PM +0800, Ming Lei wrote: > > Currently all parameters in 'affd' are read-only, so 'affd' is marked > > as const in both pci_alloc_irq_vectors_affinity() and > > irq_create_affinity_masks(). > > s/all parameters in 'affd

Re: [PATCH V3 1/5] genirq/affinity: don't mark 'affd' as const

2019-02-13 Thread Bjorn Helgaas
On Wed, Feb 13, 2019 at 06:50:37PM +0800, Ming Lei wrote: > Currently all parameters in 'affd' are read-only, so 'affd' is marked > as const in both pci_alloc_irq_vectors_affinity() and > irq_create_affinity_masks(). s/all parameters in 'affd'/the contents of '*affd'/ > We have to ask driver to

[PATCH V3 1/5] genirq/affinity: don't mark 'affd' as const

2019-02-13 Thread Ming Lei
Currently all parameters in 'affd' are read-only, so 'affd' is marked as const in both pci_alloc_irq_vectors_affinity() and irq_create_affinity_masks(). We have to ask driver to re-caculate set vectors after the whole IRQ vectors are allocated later, and the result needs to be stored in 'affd'. A