On Wed, Sep 29, 2021 at 12:50:15PM +0100, Joao Martins wrote:
> On 9/28/21 19:01, Jason Gunthorpe wrote:
> > On Thu, Sep 23, 2021 at 05:51:04PM +0100, Joao Martins wrote:
> >> So ... if pgmap accounting was removed from gup-fast then this patch
> >> would be a lot simpler and we could perhaps just
On Thu, Sep 30, 2021 at 01:01:14PM +1000, Alistair Popple wrote:
> On Thursday, 30 September 2021 5:34:05 AM AEST Jason Gunthorpe wrote:
> > On Wed, Sep 29, 2021 at 12:50:15PM +0100, Joao Martins wrote:
> >
> > > > If the get_dev_pagemap has to remain then it just means we have to
> > > > flush be
On Thu, Oct 14, 2021 at 06:56:51PM +0100, Joao Martins wrote:
> > And I would document this relationship in the GUP side "This do/while
> > is required because insert_pfn_pmd/pud() is used with compound pages
> > smaller than the PUD/PMD size" so it isn't so confused with just
> > "devmap"
>
> Al
On 10/13/21 20:43, Jason Gunthorpe wrote:
> On Wed, Oct 13, 2021 at 08:18:08PM +0100, Joao Martins wrote:
>> On 10/13/21 18:41, Jason Gunthorpe wrote:
>>> On Mon, Oct 11, 2021 at 04:53:29PM +0100, Joao Martins wrote:
On 10/8/21 12:54, Jason Gunthorpe wrote:
>>>
> The only optimization that
On Wed, Oct 13, 2021 at 08:18:08PM +0100, Joao Martins wrote:
> On 10/13/21 18:41, Jason Gunthorpe wrote:
> > On Mon, Oct 11, 2021 at 04:53:29PM +0100, Joao Martins wrote:
> >> On 10/8/21 12:54, Jason Gunthorpe wrote:
> >
> >>> The only optimization that might work here is to grab the head, then
>
On 10/13/21 18:41, Jason Gunthorpe wrote:
> On Mon, Oct 11, 2021 at 04:53:29PM +0100, Joao Martins wrote:
>> On 10/8/21 12:54, Jason Gunthorpe wrote:
>
>>> The only optimization that might work here is to grab the head, then
>>> compute the extent of tail pages and amalgamate them. Holding a ref o
On Mon, Oct 11, 2021 at 04:53:29PM +0100, Joao Martins wrote:
> On 10/8/21 12:54, Jason Gunthorpe wrote:
> > The only optimization that might work here is to grab the head, then
> > compute the extent of tail pages and amalgamate them. Holding a ref on
> > the head also secures the tails.
>
> How
On 10/8/21 12:54, Jason Gunthorpe wrote:
> On Fri, Aug 27, 2021 at 03:58:13PM +0100, Joao Martins wrote:
>> @@ -2252,16 +2265,25 @@ static int __gup_device_huge(unsigned long pfn,
>> unsigned long addr,
>> ret = 0;
>> break;
>> }
>> -
On Fri, Aug 27, 2021 at 03:58:13PM +0100, Joao Martins wrote:
> @@ -2252,16 +2265,25 @@ static int __gup_device_huge(unsigned long pfn,
> unsigned long addr,
> ret = 0;
> break;
> }
> - SetPageReferenced(page);
> - p
On Thu, Sep 30, 2021 at 06:54:05PM +0100, Joao Martins wrote:
> On 9/30/21 04:01, Alistair Popple wrote:
> > On Thursday, 30 September 2021 5:34:05 AM AEST Jason Gunthorpe wrote:
> >> On Wed, Sep 29, 2021 at 12:50:15PM +0100, Joao Martins wrote:
> >>
> If the get_dev_pagemap has to remain then
On 9/30/21 04:01, Alistair Popple wrote:
> On Thursday, 30 September 2021 5:34:05 AM AEST Jason Gunthorpe wrote:
>> On Wed, Sep 29, 2021 at 12:50:15PM +0100, Joao Martins wrote:
>>
If the get_dev_pagemap has to remain then it just means we have to
flush before changing pagemap pointers
>>
On Thursday, 30 September 2021 5:34:05 AM AEST Jason Gunthorpe wrote:
> On Wed, Sep 29, 2021 at 12:50:15PM +0100, Joao Martins wrote:
>
> > > If the get_dev_pagemap has to remain then it just means we have to
> > > flush before changing pagemap pointers
> > Right -- I don't think we should need it
On Wed, Sep 29, 2021 at 12:50:15PM +0100, Joao Martins wrote:
> > If the get_dev_pagemap has to remain then it just means we have to
> > flush before changing pagemap pointers
> Right -- I don't think we should need it as that discussion on the other
> thread goes.
>
> OTOH, using @pgmap might be
On 9/28/21 19:01, Jason Gunthorpe wrote:
> On Thu, Sep 23, 2021 at 05:51:04PM +0100, Joao Martins wrote:
>> So ... if pgmap accounting was removed from gup-fast then this patch
>> would be a lot simpler and we could perhaps just fallback to the regular
>> hugepage case (THP, HugeTLB) like your sugg
On Thu, Sep 23, 2021 at 05:51:04PM +0100, Joao Martins wrote:
> On 8/31/21 6:05 PM, Jason Gunthorpe wrote:
> >> Switching to similar iteration logic to unpin would look something like
> >> this (still untested):
> >>
> >> for_each_compound_range(index, &page, npages, head, refs) {
> >>
On 8/31/21 6:05 PM, Jason Gunthorpe wrote:
> On Tue, Aug 31, 2021 at 01:34:04PM +0100, Joao Martins wrote:
>> On 8/30/21 2:07 PM, Jason Gunthorpe wrote:
>>> On Fri, Aug 27, 2021 at 07:34:54PM +0100, Joao Martins wrote:
On 8/27/21 5:25 PM, Jason Gunthorpe wrote:
> On Fri, Aug 27, 2021 at 03
On Tue, Aug 31, 2021 at 01:34:04PM +0100, Joao Martins wrote:
> On 8/30/21 2:07 PM, Jason Gunthorpe wrote:
> > On Fri, Aug 27, 2021 at 07:34:54PM +0100, Joao Martins wrote:
> >> On 8/27/21 5:25 PM, Jason Gunthorpe wrote:
> >>> On Fri, Aug 27, 2021 at 03:58:13PM +0100, Joao Martins wrote:
> >>>
> >>
On 8/30/21 2:07 PM, Jason Gunthorpe wrote:
> On Fri, Aug 27, 2021 at 07:34:54PM +0100, Joao Martins wrote:
>> On 8/27/21 5:25 PM, Jason Gunthorpe wrote:
>>> On Fri, Aug 27, 2021 at 03:58:13PM +0100, Joao Martins wrote:
>>>
#if defined(CONFIG_ARCH_HAS_PTE_DEVMAP) &&
defined(CONFIG_TRANSP
On Fri, Aug 27, 2021 at 07:34:54PM +0100, Joao Martins wrote:
> On 8/27/21 5:25 PM, Jason Gunthorpe wrote:
> > On Fri, Aug 27, 2021 at 03:58:13PM +0100, Joao Martins wrote:
> >
> >> #if defined(CONFIG_ARCH_HAS_PTE_DEVMAP) &&
> >> defined(CONFIG_TRANSPARENT_HUGEPAGE)
> >> static int __gup_device
On 8/27/21 5:25 PM, Jason Gunthorpe wrote:
> On Fri, Aug 27, 2021 at 03:58:13PM +0100, Joao Martins wrote:
>
>> #if defined(CONFIG_ARCH_HAS_PTE_DEVMAP) &&
>> defined(CONFIG_TRANSPARENT_HUGEPAGE)
>> static int __gup_device_huge(unsigned long pfn, unsigned long addr,
>>
On Fri, Aug 27, 2021 at 03:58:13PM +0100, Joao Martins wrote:
> #if defined(CONFIG_ARCH_HAS_PTE_DEVMAP) &&
> defined(CONFIG_TRANSPARENT_HUGEPAGE)
> static int __gup_device_huge(unsigned long pfn, unsigned long addr,
>unsigned long end, unsigned int flags,
>
Use try_grab_compound_head() for device-dax GUP when configured with a
compound devmap.
Rather than incrementing the refcount for each page, do one atomic
addition for all the pages to be pinned.
Performance measured by gup_benchmark improves considerably
get_user_pages_fast() and pin_user_pages_
22 matches
Mail list logo
