Re: [PATCH] arm64: cache: Lower ARCH_DMA_MINALIGN to 64 (L1_CACHE_BYTES)

On Wed, Jun 2, 2021 at 4:28 PM Marek Szyprowski
[off-list ref] wrote:

On 02.06.2021 16:14, Arnd Bergmann wrote:

quoted

On Wed, Jun 2, 2021 at 4:09 PM Marek Szyprowski
[off-list ref] wrote:

quoted

On 02.06.2021 15:51, Mark Rutland wrote:
[   13.938691] debugfs: Directory '3f902000.hdmi' with parent 'vc4-hdmi'
already present!
[   13.983716] vc4-drm soc:gpu: bound 3f902000.hdmi (ops vc4_hdmi_ops)
[   13.991816] vc4-drm soc:gpu: bound 3f806000.vec (ops vc4_vec_ops)
[   14.000945] vc4-drm soc:gpu: bound 3f400000.hvs (ops vc4_hvs_ops)
[   14.009644] vc4-drm soc:gpu: bound 3f004000.txp (ops vc4_txp_ops)
[   14.017547] vc4-drm soc:gpu: bound 3f206000.pixelvalve (ops vc4_crtc_ops)
[   14.025780] vc4-drm soc:gpu: bound 3f207000.pixelvalve (ops vc4_crtc_ops)
[   14.033995] vc4-drm soc:gpu: bound 3f807000.pixelvalve (ops vc4_crtc_ops)
[   14.042196] vc4-drm soc:gpu: bound 3fc00000.v3d (ops vc4_v3d_ops)
[....] Waiting for /dev to be fully populated...[   14.112812] [drm]
Initialized vc4 0.0.0 20140616 for soc:gpu on minor 0
done.

I wonder if the vc4's cache is involved here rather than the CPU cache.
If it's that, then removing the vc4 drivers should make it boot reliably, and
it might be possible to fix it by manually padding any allocations that get
passed to the GPU.

Indeed, after disabling DRM_VC4 in the .config, system boots fine and
seems to be working correctly.

Ok, that helps. This means there is a good chance it would be one of these:

$ git grep k[cmz]alloc drivers/gpu/drm/vc4
drivers/gpu/drm/vc4/vc4_bo.c:           new_list =
kmalloc_array(new_size, sizeof(struct list_head),
drivers/gpu/drm/vc4/vc4_bo.c:   bo = kzalloc(sizeof(*bo), GFP_KERNEL);
drivers/gpu/drm/vc4/vc4_bo.c:   vc4->bo_labels =
kcalloc(VC4_BO_TYPE_COUNT, sizeof(*vc4->bo_labels),
drivers/gpu/drm/vc4/vc4_crtc.c: flip_state =
kzalloc(sizeof(*flip_state), GFP_KERNEL);
drivers/gpu/drm/vc4/vc4_crtc.c: vc4_state =
kzalloc(sizeof(*vc4_state), GFP_KERNEL);
drivers/gpu/drm/vc4/vc4_crtc.c: vc4_crtc_state =
kzalloc(sizeof(*vc4_crtc_state), GFP_KERNEL);
drivers/gpu/drm/vc4/vc4_crtc.c: vc4_crtc = devm_kzalloc(dev,
sizeof(*vc4_crtc), GFP_KERNEL);
drivers/gpu/drm/vc4/vc4_debugfs.c:              devm_kzalloc(dev->dev,
sizeof(*entry), GFP_KERNEL);
drivers/gpu/drm/vc4/vc4_dpi.c:  dpi = devm_kzalloc(dev, sizeof(*dpi),
GFP_KERNEL);
drivers/gpu/drm/vc4/vc4_dpi.c:  vc4_dpi_encoder = devm_kzalloc(dev,
sizeof(*vc4_dpi_encoder),
drivers/gpu/drm/vc4/vc4_drv.c:  vc4file = kzalloc(sizeof(*vc4file), GFP_KERNEL);
drivers/gpu/drm/vc4/vc4_dsi.c:  dsi->clk_onecell = devm_kzalloc(dev,
drivers/gpu/drm/vc4/vc4_dsi.c:  vc4_dsi_encoder = devm_kzalloc(dev,
sizeof(*vc4_dsi_encoder),
drivers/gpu/drm/vc4/vc4_dsi.c:  dsi = devm_kzalloc(dev, sizeof(*dsi),
GFP_KERNEL);
drivers/gpu/drm/vc4/vc4_gem.c:  bo_state = kcalloc(state->bo_count,
sizeof(*bo_state), GFP_KERNEL);
drivers/gpu/drm/vc4/vc4_gem.c:  kernel_state = kcalloc(1,
sizeof(*kernel_state), GFP_KERNEL);
drivers/gpu/drm/vc4/vc4_gem.c:  kernel_state->bo = kcalloc(state->bo_count,
drivers/gpu/drm/vc4/vc4_gem.c:  fence = kzalloc(sizeof(*fence), GFP_KERNEL);
drivers/gpu/drm/vc4/vc4_gem.c:  exec = kcalloc(1, sizeof(*exec), GFP_KERNEL);
drivers/gpu/drm/vc4/vc4_hdmi.c:         kzalloc(sizeof(*new_state), GFP_KERNEL);
drivers/gpu/drm/vc4/vc4_hdmi.c: new_state =
kzalloc(sizeof(*new_state), GFP_KERNEL);
drivers/gpu/drm/vc4/vc4_hdmi.c: regs = kcalloc(variant->num_registers,
sizeof(*regs),
drivers/gpu/drm/vc4/vc4_hdmi.c: vc4_hdmi = devm_kzalloc(dev,
sizeof(*vc4_hdmi), GFP_KERNEL);
drivers/gpu/drm/vc4/vc4_hvs.c:  hvs = devm_kzalloc(&pdev->dev,
sizeof(*hvs), GFP_KERNEL);
drivers/gpu/drm/vc4/vc4_kms.c:  ctm_state =
kzalloc(sizeof(*ctm_state), GFP_KERNEL);
drivers/gpu/drm/vc4/vc4_kms.c:  load_state =
kzalloc(sizeof(*load_state), GFP_KERNEL);
drivers/gpu/drm/vc4/vc4_kms.c:  state = kzalloc(sizeof(*state), GFP_KERNEL);
drivers/gpu/drm/vc4/vc4_kms.c:  state = kzalloc(sizeof(*state), GFP_KERNEL);
drivers/gpu/drm/vc4/vc4_perfmon.c:      perfmon =
kzalloc(struct_size(perfmon, counters, req->ncounters),
drivers/gpu/drm/vc4/vc4_plane.c:        vc4_state =
kzalloc(sizeof(*vc4_state), GFP_KERNEL);
drivers/gpu/drm/vc4/vc4_plane.c:                u32 *new_dlist =
kmalloc_array(new_size, 4, GFP_KERNEL);
drivers/gpu/drm/vc4/vc4_plane.c:        vc4_plane =
devm_kzalloc(dev->dev, sizeof(*vc4_plane),
drivers/gpu/drm/vc4/vc4_txp.c:  txp = devm_kzalloc(dev, sizeof(*txp),
GFP_KERNEL);
drivers/gpu/drm/vc4/vc4_v3d.c:  v3d = devm_kzalloc(&pdev->dev,
sizeof(*v3d), GFP_KERNEL);
drivers/gpu/drm/vc4/vc4_validate_shaders.c:
kcalloc(BITS_TO_LONGS(validation_state.max_ip),
drivers/gpu/drm/vc4/vc4_validate_shaders.c:     validated_shader =
kcalloc(1, sizeof(*validated_shader), GFP_KERNEL);
drivers/gpu/drm/vc4/vc4_vec.c:  vec_connector = devm_kzalloc(dev->dev,
sizeof(*vec_connector),
drivers/gpu/drm/vc4/vc4_vec.c:  vec = devm_kzalloc(dev, sizeof(*vec),
GFP_KERNEL);
drivers/gpu/drm/vc4/vc4_vec.c:  vc4_vec_encoder = devm_kzalloc(dev,
sizeof(*vc4_vec_encoder),

I suppose most can be easily ruled out because they are not shared, or
because they are larger
than 64 bytes. I'll have a quick look if I find any smoking guns.

       Arnd

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel