Re: [PATCH] mm/gup: restore the ability to pin more than 2GB at a time

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

To: David Hildenbrand <david@xxxxxxxxxx>
Subject: Re: [PATCH] mm/gup: restore the ability to pin more than 2GB at a time
From: Jason Gunthorpe <jgg@xxxxxxxxxx>
Date: Wed, 30 Oct 2024 09:04:53 -0300
Cc: John Hubbard <jhubbard@xxxxxxxxxx>, Alistair Popple <apopple@xxxxxxxxxx>, Christoph Hellwig <hch@xxxxxxxxxxxxx>, Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>, LKML <linux-kernel@xxxxxxxxxxxxxxx>, linux-mm@xxxxxxxxx, linux-stable@xxxxxxxxxxxxxxx, Vivek Kasireddy <vivek.kasireddy@xxxxxxxxx>, Dave Airlie <airlied@xxxxxxxxxx>, Gerd Hoffmann <kraxel@xxxxxxxxxx>, Matthew Wilcox <willy@xxxxxxxxxxxxx>, Peter Xu <peterx@xxxxxxxxxx>, Arnd Bergmann <arnd@xxxxxxxx>, Daniel Vetter <daniel.vetter@xxxxxxxx>, Dongwon Kim <dongwon.kim@xxxxxxxxx>, Hugh Dickins <hughd@xxxxxxxxxx>, Junxiao Chang <junxiao.chang@xxxxxxxxx>, Mike Kravetz <mike.kravetz@xxxxxxxxxx>, Oscar Salvador <osalvador@xxxxxxx>
In-reply-to: <bfee966f-807d-4668-b353-159a6e8066f2@redhat.com>
References: <20241030030116.670307-1-jhubbard@nvidia.com> <ZyG0VKUpFttPF30f@infradead.org> <249d2614-0bcc-4ca8-b24e-7c0578a81dce@nvidia.com> <ZyG3GAvTHpRL9tnU@infradead.org> <ea81f12a-95a3-4b9d-90e7-53a5d9c910be@nvidia.com> <ZyG5IumNPMUDBQOq@infradead.org> <fa766610-4a0c-4d75-90fd-6c781fadee73@nvidia.com> <87r07yp0ng.fsf@nvdebian.thelocal> <128d04dd-2d48-4a98-8537-49589b4db1c3@nvidia.com> <bfee966f-807d-4668-b353-159a6e8066f2@redhat.com>

On Wed, Oct 30, 2024 at 09:34:51AM +0100, David Hildenbrand wrote:

> The unusual thing is not the amount of system memory we are pinning but *how
> many* pages we try pinning in the single call.
> 
> If you stare at vfio_pin_pages_remote, we seem to be batching it.
> 
> long req_pages = min_t(long, npage, batch->capacity);
> 
> Which is
> 
> #define VFIO_BATCH_MAX_CAPACITY (PAGE_SIZE / sizeof(struct page *))
> 
> So you can fix this in your driver ;)

Yeah, everything batches that I'm aware of. RDMA also uses a 4k batch
size, and iommufd uses 64k.

Jason

Follow-Ups:
- Re: [PATCH] mm/gup: restore the ability to pin more than 2GB at a time
  - From: John Hubbard

References:
- [PATCH] mm/gup: restore the ability to pin more than 2GB at a time
  - From: John Hubbard
- Re: [PATCH] mm/gup: restore the ability to pin more than 2GB at a time
  - From: Christoph Hellwig
- Re: [PATCH] mm/gup: restore the ability to pin more than 2GB at a time
  - From: John Hubbard
- Re: [PATCH] mm/gup: restore the ability to pin more than 2GB at a time
  - From: Christoph Hellwig
- Re: [PATCH] mm/gup: restore the ability to pin more than 2GB at a time
  - From: John Hubbard
- Re: [PATCH] mm/gup: restore the ability to pin more than 2GB at a time
  - From: Christoph Hellwig
- Re: [PATCH] mm/gup: restore the ability to pin more than 2GB at a time
  - From: John Hubbard
- Re: [PATCH] mm/gup: restore the ability to pin more than 2GB at a time
  - From: Alistair Popple
- Re: [PATCH] mm/gup: restore the ability to pin more than 2GB at a time
  - From: John Hubbard
- Re: [PATCH] mm/gup: restore the ability to pin more than 2GB at a time
  - From: David Hildenbrand

Prev by Date: Re: [PATCH hotfix 6.12 v4 4/5] mm: refactor arch_calc_vm_flag_bits() and arm64 MTE handling
Next by Date: [PATCH v7 15/32] riscv/shstk: If needed allocate a new shadow stack on clone
Previous by thread: Re: [PATCH] mm/gup: restore the ability to pin more than 2GB at a time
Next by thread: Re: [PATCH] mm/gup: restore the ability to pin more than 2GB at a time
Index(es):
- Date
- Thread

[Index of Archives] [Linux ARM Kernel] [Linux ARM] [Linux Omap] [Fedora ARM] [IETF Annouce] [Bugtraq] [Linux OMAP] [Linux MIPS] [eCos] [Asterisk Internet PBX] [Linux API]