linux-kernel - Re: [PATCH v2 1/6] drm/v3d: Document cache flushing ABI.

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CAApk1dWeKKrk1SpOQFi_6VLjfE-+4NztBd3RULhhwubzCtjqOw@mail.gmail.com>
Date:   Tue, 4 Dec 2018 11:51:27 +0000
From:   Dave Emett <david.emett@...adcom.com>
To:     Eric Anholt <eric@...olt.net>
Cc:     dri-devel@...ts.freedesktop.org, linux-kernel@...r.kernel.org,
        Thomas Spurden <thomas.spurden@...adcom.com>
Subject: Re: [PATCH v2 1/6] drm/v3d: Document cache flushing ABI.

On Mon, 3 Dec 2018 at 22:24, Eric Anholt <eric@...olt.net> wrote:
>
> Right now, userspace doesn't do any L2T writes, but we should lay out
> our expectations for how it works.
>
> v2: Explicitly mention the VCD cache flushing requirements and that
>     we'll flush the other caches before each of the CLs.
>
> Signed-off-by: Eric Anholt <eric@...olt.net>

Reviewed-by: Dave Emett <david.emett@...adcom.com>

> ---
>  include/uapi/drm/v3d_drm.h | 8 ++++++++
>  1 file changed, 8 insertions(+)
>
> diff --git a/include/uapi/drm/v3d_drm.h b/include/uapi/drm/v3d_drm.h
> index 35c7d813c66e..ea70669d2138 100644
> --- a/include/uapi/drm/v3d_drm.h
> +++ b/include/uapi/drm/v3d_drm.h
> @@ -52,6 +52,14 @@ extern "C" {
>   *
>   * This asks the kernel to have the GPU execute an optional binner
>   * command list, and a render command list.
> + *
> + * The L1T, slice, L2C, L2T, and GCA caches will be flushed before
> + * each CL executes.  The VCD cache should be flushed (if necessary)
> + * by the submitted CLs.  The TLB writes are guaranteed to have been
> + * flushed by the time the render done IRQ happens, which is the
> + * trigger for out_sync.  Any dirtying of cachelines by the job (only
> + * possible using TMU writes) must be flushed by the caller using the
> + * CL's cache flush commands.
>   */
>  struct drm_v3d_submit_cl {
>         /* Pointer to the binner command list.
> --
> 2.20.0.rc1
>