public inbox for igt-dev@lists.freedesktop.org
 help / color / mirror / Atom feed
From: "Hellstrom, Thomas" <thomas.hellstrom@intel.com>
To: "igt-dev@lists.freedesktop.org" <igt-dev@lists.freedesktop.org>,
	"Sharma,  Nishit" <nishit.sharma@intel.com>
Subject: Re: [PATCH i-g-t 1/2] lib/xe: Introduce aligned buffer mapping
Date: Wed, 11 Mar 2026 18:53:37 +0000	[thread overview]
Message-ID: <aeee76de85ea852c8ab6f8e886a0046fb3bb6dec.camel@intel.com> (raw)
In-Reply-To: <20260311170354.475828-2-nishit.sharma@intel.com>

On Wed, 2026-03-11 at 17:03 +0000, nishit.sharma@intel.com wrote:
> From: Nishit Sharma <nishit.sharma@intel.com>
> 
> Added aligned buffer mapping which is providing an interface for
> mapping
> buffer objects with a specified alignment. This API improves
> cross-platform compatibility and simplifies batch buffer setup by
> eliminating the need for platform-specific address handling.
> 
> Signed-off-by: Thomas Hellstrom <thomas.hellstrom@intel.com>
> Signed-off-by: Nishit Sharma <nishit.sharma@intel.com>
> ---
>  lib/xe/xe_ioctl.c | 40 ++++++++++++++++++++++++++++++++++++++++
>  lib/xe/xe_ioctl.h |  1 +
>  2 files changed, 41 insertions(+)
> 
> diff --git a/lib/xe/xe_ioctl.c b/lib/xe/xe_ioctl.c
> index 16aae05c9..2e9e75656 100644
> --- a/lib/xe/xe_ioctl.c
> +++ b/lib/xe/xe_ioctl.c
> @@ -579,6 +579,46 @@ void *xe_bo_map_fixed(int fd, uint32_t bo,
> size_t size, uint64_t addr)
>  	return map;
>  }
>  
> +/**
> + * xe_bo_map_aligned() - Create a bo CPU map requiring an alignment
> for the
> + * returned address.
> + * @fd: The device file-descriptor
> + * @bo: The buffer object
> + * @size: The size of the map
> + * @alignment: The requested map alingment.

s/alingment/alignment/

Return:

> + */
> +void *xe_bo_map_aligned(int fd, uint32_t bo, size_t size, size_t
> alignment)
> +{
> +	size_t anon_size = size + alignment;
> +	uint64_t anon_addr;
> +	void *anon_map, *map;
> +	uint64_t map_addr;
> +	size_t hole_size;
> +
> +
> +	/* Reserve a range of virtual space where we can fit an
> aligned map */
> +	anon_map = mmap(NULL, anon_size, PROT_NONE, MAP_PRIVATE |
> MAP_ANONYMOUS, -1, 0);
> +	igt_assert(anon_map != MAP_FAILED);
> +	anon_addr = to_user_pointer(anon_map);
> +
> +	/* Compute the first aligned address within the virtual
> space. */
> +	map_addr = ALIGN(anon_addr, alignment);
> +	/* Map the bo there, replacing part of the reserved virtual
> range. */
> +	map = xe_bo_map_fixed(fd, bo, size, map_addr);

Perhaps an igt_assert() that map is indeed properly aligned.

> +
> +	/* Unreserve part of the virtual range (if any) *before* the
> bo map */
> +	hole_size = map_addr - anon_addr;
> +	if (hole_size)
> +		igt_assert(munmap(anon_map, hole_size) == 0);
> +
> +	/* Unreserve part of the virtual range (if any) *after* the
> bo map */
> +	hole_size = anon_size - hole_size - size;
> +	if (hole_size)
> +		igt_assert(munmap(map + size, hole_size) == 0);
> +
> +	return map;
> +}
> +
>  void *xe_bo_mmap_ext(int fd, uint32_t bo, size_t size, int prot)
>  {
>  	return __xe_bo_map(fd, bo, size, prot);
> diff --git a/lib/xe/xe_ioctl.h b/lib/xe/xe_ioctl.h
> index 3ea651063..b62d259fd 100644
> --- a/lib/xe/xe_ioctl.h
> +++ b/lib/xe/xe_ioctl.h
> @@ -91,6 +91,7 @@ void xe_exec_queue_destroy(int fd, uint32_t
> exec_queue);
>  uint64_t xe_bo_mmap_offset(int fd, uint32_t bo);
>  void *xe_bo_map(int fd, uint32_t bo, size_t size);
>  void *xe_bo_map_fixed(int fd, uint32_t bo, size_t size, uint64_t
> addr);
> +void *xe_bo_map_aligned(int fd, uint32_t bo, size_t size, size_t
> alignment);
>  void *xe_bo_mmap_ext(int fd, uint32_t bo, size_t size, int prot);
>  int __xe_exec(int fd, struct drm_xe_exec *exec);
>  void xe_exec(int fd, struct drm_xe_exec *exec);

Otherwise need an additional reviewer, since I contributed to the
patch.

Thanks,
Thomas



  reply	other threads:[~2026-03-11 18:53 UTC|newest]

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-03-11 17:03 [PATCH i-g-t 0/2] Unify Batch Buffer Alignment and Introduce Aligned Buffer Mapping nishit.sharma
2026-03-11 17:03 ` [PATCH i-g-t 1/2] lib/xe: Introduce aligned buffer mapping nishit.sharma
2026-03-11 18:53   ` Hellstrom, Thomas [this message]
2026-03-11 17:03 ` [PATCH i-g-t 2/2] tests/intel/xe_svm_usrptr_madvise: Unify batch buffer alignment nishit.sharma
2026-03-11 18:54   ` Hellstrom, Thomas
2026-03-12  1:39 ` ✓ Xe.CI.BAT: success for Unify Batch Buffer Alignment and Introduce Aligned Buffer Mapping Patchwork
2026-03-12  1:48 ` ✓ i915.CI.BAT: " Patchwork
2026-03-12 21:03 ` ✗ Xe.CI.FULL: failure " Patchwork
2026-03-12 21:45 ` ✓ i915.CI.Full: success " Patchwork

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=aeee76de85ea852c8ab6f8e886a0046fb3bb6dec.camel@intel.com \
    --to=thomas.hellstrom@intel.com \
    --cc=igt-dev@lists.freedesktop.org \
    --cc=nishit.sharma@intel.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox