All of lore.kernel.org
 help / color / mirror / Atom feed
From: sashiko-bot@kernel.org
To: "Rob Herring (Arm)" <robh@kernel.org>
Cc: dri-devel@lists.freedesktop.org
Subject: Re: [PATCH v5 14/19] accel: ethosu: Validate OFM transpose
Date: Fri, 20 Nov 2026 15:07:03 +0000	[thread overview]
Message-ID: <sashiko-outbox-225427@kernel.org> (raw)
In-Reply-To: <20261120-ethosu-fixes-v5-14-52d459acdb84@kernel.org>

Thank you for your contribution! Sashiko AI review found 1 potential
issue to consider:
- [Critical] accel: ethosu: missing NHCWB16 format check when OFM
  transpose is set
--

commit 819bad25b9dbcbe54db35bf0c5afd5f5f47c3c99
Author: Rob Herring (Arm) <robh@kernel.org>

accel: ethosu: Validate OFM transpose

This commit adds validation for Ethos-U85 output feature map (OFM)
transpose encodings and permutes the OFM dimensions before validating
tile and stride accesses.

> diff --git a/drivers/accel/ethosu/ethosu_gem.c b/drivers/accel/ethosu/ethosu_gem.c
> index 0ab73b99c458c..b8f840ab1ceed 100644
> --- a/drivers/accel/ethosu/ethosu_gem.c
> +++ b/drivers/accel/ethosu/ethosu_gem.c

[ ... ]

> @@ -1029,8 +1083,19 @@ static int ethosu_gem_cmdstream_copy_and_validate(struct drm_device *ddev,
>  		case NPU_SET_OFM_PRECISION:
>  			if (((param >> 6) & 0x3) > 1)
>  				return -EINVAL;
> -			if (!ethosu_is_u65(edev) && (param & GENMASK(13, 11)))
> -				return -EINVAL;
> +			if (!ethosu_is_u65(edev)) {
> +				u32 transpose = FIELD_GET(NPU_OFM_TRANSPOSE_MASK, param);
> +
> +				switch (transpose) {
> +				case 4:
> +				case 5:
> +					return -EINVAL;
> +				}
> +				if (transpose && ((param >> 1) & 0x3) == 3)
> +					return -EINVAL;
> +				if ((param >> 14) == 2 && transpose)
> +					return -EINVAL;
> +			}
>  			st.ofm.precision = param;
>  			break;
>  		case NPU_SET_OFM_REGION:

[Severity: Critical]
Should NPU_SET_OFM_PRECISION also reject a non-zero transpose when the
activation format is NHCWB16 (((param >> 6) & 0x3) == 1)?

OFM transposition on U85 is only valid with the linear NHWC format.  If
NHCWB16 is combined with a non-zero transpose, feat_matrix_validate() only
requires NPU_SET_OFM_STRIDE_C to be set, and feat_matrix_length() computes
the NHCWB16 brick offset on the coordinates already permuted by
feat_matrix_permute() while ignoring fm->stride_x.  That can underestimate
info->region_size[fm->region] and allow out-of-bounds NPU DMA writes past
the end of the GEM buffer.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261120-ethosu-fixes-v5-0-52d459acdb84@kernel.org?part=14

  reply	other threads:[~2026-11-20 15:07 UTC|newest]

Thread overview: 22+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-20 14:57 [PATCH v5 00/19] accel: ethosu: Cmd stream validation fixes Rob Herring (Arm)
2026-11-20 14:57 ` [PATCH v5 01/19] accel: ethosu: Prevent command stream export Rob Herring (Arm)
2026-11-20 14:57 ` [PATCH v5 02/19] accel: ethosu: Move DMA mode to src/dst struct Rob Herring (Arm)
2026-11-20 14:57 ` [PATCH v5 03/19] accel: ethosu: Track command stream register setup Rob Herring (Arm)
2026-11-20 14:57 ` [PATCH v5 04/19] accel: ethosu: Factor buffer bounds checks Rob Herring (Arm)
2026-11-20 14:57 ` [PATCH v5 05/19] accel: ethosu: Fix NHCWB16 bounds calculation Rob Herring (Arm)
2026-11-20 14:57 ` [PATCH v5 06/19] accel: ethosu: Validate secondary streams Rob Herring (Arm)
2026-11-20 14:57 ` [PATCH v5 07/19] accel: ethosu: Reject unsupported commands and params Rob Herring (Arm)
2026-11-20 14:57 ` [PATCH v5 08/19] accel: ethosu: Validate all feature map tiles Rob Herring (Arm)
2026-11-20 14:57 ` [PATCH v5 09/19] accel: ethosu: Account for feature map element size Rob Herring (Arm)
2026-11-20 14:57 ` [PATCH v5 10/19] accel: ethosu: Validate convolution parameter Rob Herring (Arm)
2026-11-20 14:57 ` [PATCH v5 11/19] accel: ethosu: Validate accumulator input Rob Herring (Arm)
2026-11-20 14:57 ` [PATCH v5 12/19] accel: ethosu: Restrict dynamic IFM2 weights Rob Herring (Arm)
2026-11-20 14:57 ` [PATCH v5 13/19] accel: ethosu: Split U65 and U85 DMA length validation Rob Herring (Arm)
2026-11-20 14:57 ` [PATCH v5 14/19] accel: ethosu: Validate OFM transpose Rob Herring (Arm)
2026-11-20 15:07   ` sashiko-bot [this message]
2026-11-20 14:57 ` [PATCH v5 15/19] accel: ethosu: Validate resize operations Rob Herring (Arm)
2026-11-20 14:57 ` [PATCH v5 16/19] accel: ethosu: Ensure cmd stream is multiple of 4 bytes Rob Herring (Arm)
2026-11-20 14:57 ` [PATCH v5 17/19] accel: ethosu: Require active weight streams setup Rob Herring (Arm)
2026-11-20 14:57 ` [PATCH v5 18/19] accel: ethosu: Support IFM upscaling Rob Herring (Arm)
2026-11-20 15:12   ` sashiko-bot
2026-11-20 14:57 ` [PATCH v5 19/19] accel: ethosu: Reject IFM broadcast on U65 Rob Herring (Arm)

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=sashiko-outbox-225427@kernel.org \
    --to=sashiko-bot@kernel.org \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=robh@kernel.org \
    --cc=sashiko-reviews@lists.linux.dev \
    /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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.