From 211ecfaaef186ee5230a77d054cdec7fbfc6724a Mon Sep 17 00:00:00 2001 From: Brad Spengler Date: Wed, 7 Jan 2026 12:12:36 -0500 Subject: [PATCH 01/25] drm/vmwgfx: Fix invalid kref_put callback in vmw_bo_dirty_release The kref_put() call uses (void *)kvfree as the release callback, which is incorrect. kref_put() expects a function with signature void (*release)(struct kref *), but kvfree has signature void (*)(const void *). Calling through an incompatible function pointer is undefined behavior. The code only worked by accident because ref_count is the first member of vmw_bo_dirty, making the kref pointer equal to the struct pointer. Fix this by adding a proper release callback that uses container_of() to retrieve the containing structure before freeing. Fixes: c1962742ffff ("drm/vmwgfx: Use kref in vmw_bo_dirty") Signed-off-by: Brad Spengler Signed-off-by: Zack Rusin Cc: Ian Forbes Link: https://patch.msgid.link/20260107171236.3573118-1-zack.rusin@broadcom.com --- drivers/gpu/drm/vmwgfx/vmwgfx_page_dirty.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_page_dirty.c b/drivers/gpu/drm/vmwgfx/vmwgfx_page_dirty.c index fd4e76486f2d..45561bc1c9ef 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_page_dirty.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_page_dirty.c @@ -260,6 +260,13 @@ int vmw_bo_dirty_add(struct vmw_bo *vbo) return ret; } +static void vmw_bo_dirty_free(struct kref *kref) +{ + struct vmw_bo_dirty *dirty = container_of(kref, struct vmw_bo_dirty, ref_count); + + kvfree(dirty); +} + /** * vmw_bo_dirty_release - Release a dirty-tracking user from a buffer object * @vbo: The buffer object @@ -274,7 +281,7 @@ void vmw_bo_dirty_release(struct vmw_bo *vbo) { struct vmw_bo_dirty *dirty = vbo->dirty; - if (dirty && kref_put(&dirty->ref_count, (void *)kvfree)) + if (dirty && kref_put(&dirty->ref_count, vmw_bo_dirty_free)) vbo->dirty = NULL; } From 922f9dec5d19df4cfbb7070275e5c131d10c80f3 Mon Sep 17 00:00:00 2001 From: Ian Forbes Date: Fri, 9 Jan 2026 09:51:39 -0600 Subject: [PATCH 02/25] drm/vmwgfx: Set a unique ID for each submitted command buffer These IDs are logged by the Hypervisor when debug logging is enabled. Having the IDs in the log makes it much easier to see when command buffers start and finish. They can also be used by logging/tracing in the Guest to help correlate between Guest and Hypervisor logs. Signed-off-by: Ian Forbes Signed-off-by: Zack Rusin Link: https://patch.msgid.link/20260109155139.3259493-1-ian.forbes@broadcom.com --- drivers/gpu/drm/vmwgfx/vmwgfx_cmdbuf.c | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_cmdbuf.c b/drivers/gpu/drm/vmwgfx/vmwgfx_cmdbuf.c index 94e8982f5616..1ee37690b940 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_cmdbuf.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_cmdbuf.c @@ -105,6 +105,7 @@ struct vmw_cmdbuf_context { * @handle: DMA address handle for the command buffer space if @using_mob is * false. Immutable. * @size: The size of the command buffer space. Immutable. + * @id: Monotonically increasing ID of the last cmdbuf submitted. * @num_contexts: Number of contexts actually enabled. */ struct vmw_cmdbuf_man { @@ -132,6 +133,7 @@ struct vmw_cmdbuf_man { bool has_pool; dma_addr_t handle; size_t size; + u64 id; u32 num_contexts; }; @@ -303,6 +305,8 @@ static int vmw_cmdbuf_header_submit(struct vmw_cmdbuf_header *header) struct vmw_cmdbuf_man *man = header->man; u32 val; + header->cb_header->id = man->id++; + val = upper_32_bits(header->handle); vmw_write(man->dev_priv, SVGA_REG_COMMAND_HIGH, val); From 5023ca80f9589295cb60735016e39fc5cc714243 Mon Sep 17 00:00:00 2001 From: Ian Forbes Date: Tue, 13 Jan 2026 11:53:57 -0600 Subject: [PATCH 03/25] drm/vmwgfx: Return the correct value in vmw_translate_ptr functions Before the referenced fixes these functions used a lookup function that returned a pointer. This was changed to another lookup function that returned an error code with the pointer becoming an out parameter. The error path when the lookup failed was not changed to reflect this change and the code continued to return the PTR_ERR of the now uninitialized pointer. This could cause the vmw_translate_ptr functions to return success when they actually failed causing further uninitialized and OOB accesses. Reported-by: Kuzey Arda Bulut Fixes: a309c7194e8a ("drm/vmwgfx: Remove rcu locks from user resources") Signed-off-by: Ian Forbes Reviewed-by: Zack Rusin Signed-off-by: Zack Rusin Link: https://patch.msgid.link/20260113175357.129285-1-ian.forbes@broadcom.com --- drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c b/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c index 3057f8baa7d2..e1f18020170a 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c @@ -1143,7 +1143,7 @@ static int vmw_translate_mob_ptr(struct vmw_private *dev_priv, ret = vmw_user_bo_lookup(sw_context->filp, handle, &vmw_bo); if (ret != 0) { drm_dbg(&dev_priv->drm, "Could not find or use MOB buffer.\n"); - return PTR_ERR(vmw_bo); + return ret; } vmw_bo_placement_set(vmw_bo, VMW_BO_DOMAIN_MOB, VMW_BO_DOMAIN_MOB); ret = vmw_validation_add_bo(sw_context->ctx, vmw_bo); @@ -1199,7 +1199,7 @@ static int vmw_translate_guest_ptr(struct vmw_private *dev_priv, ret = vmw_user_bo_lookup(sw_context->filp, handle, &vmw_bo); if (ret != 0) { drm_dbg(&dev_priv->drm, "Could not find or use GMR region.\n"); - return PTR_ERR(vmw_bo); + return ret; } vmw_bo_placement_set(vmw_bo, VMW_BO_DOMAIN_GMR | VMW_BO_DOMAIN_VRAM, VMW_BO_DOMAIN_GMR | VMW_BO_DOMAIN_VRAM); From fef0e649f8b42bdffe4a916dd46e1b1e9ad2f207 Mon Sep 17 00:00:00 2001 From: Felix Gu Date: Fri, 30 Jan 2026 00:21:19 +0800 Subject: [PATCH 04/25] drm/logicvc: Fix device node reference leak in logicvc_drm_config_parse() The logicvc_drm_config_parse() function calls of_get_child_by_name() to find the "layers" node but fails to release the reference, leading to a device node reference leak. Fix this by using the __free(device_node) cleanup attribute to automatic release the reference when the variable goes out of scope. Fixes: efeeaefe9be5 ("drm: Add support for the LogiCVC display controller") Signed-off-by: Felix Gu Reviewed-by: Luca Ceresoli Reviewed-by: Kory Maincent Link: https://patch.msgid.link/20260130-logicvc_drm-v1-1-04366463750c@gmail.com Signed-off-by: Luca Ceresoli --- drivers/gpu/drm/logicvc/logicvc_drm.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/logicvc/logicvc_drm.c b/drivers/gpu/drm/logicvc/logicvc_drm.c index 204b0fee55d0..bbebf4fc7f51 100644 --- a/drivers/gpu/drm/logicvc/logicvc_drm.c +++ b/drivers/gpu/drm/logicvc/logicvc_drm.c @@ -92,7 +92,6 @@ static int logicvc_drm_config_parse(struct logicvc_drm *logicvc) struct device *dev = drm_dev->dev; struct device_node *of_node = dev->of_node; struct logicvc_drm_config *config = &logicvc->config; - struct device_node *layers_node; int ret; logicvc_of_property_parse_bool(of_node, LOGICVC_OF_PROPERTY_DITHERING, @@ -128,7 +127,8 @@ static int logicvc_drm_config_parse(struct logicvc_drm *logicvc) if (ret) return ret; - layers_node = of_get_child_by_name(of_node, "layers"); + struct device_node *layers_node __free(device_node) = + of_get_child_by_name(of_node, "layers"); if (!layers_node) { drm_err(drm_dev, "Missing non-optional layers node\n"); return -EINVAL; From 0b87d51690dd5131cbe9fbd23746b037aab89815 Mon Sep 17 00:00:00 2001 From: Franz Schnyder Date: Fri, 6 Feb 2026 13:37:36 +0100 Subject: [PATCH 05/25] drm/bridge: ti-sn65dsi86: Enable HPD polling if IRQ is not used Fallback to polling to detect hotplug events on systems without interrupts. On systems where the interrupt line of the bridge is not connected, the bridge cannot notify hotplug events. Only add the DRM_BRIDGE_OP_HPD flag if an interrupt has been registered otherwise remain in polling mode. Fixes: 55e8ff842051 ("drm/bridge: ti-sn65dsi86: Add HPD for DisplayPort connector type") Cc: stable@vger.kernel.org # 6.16: 9133bc3f0564: drm/bridge: ti-sn65dsi86: Add Signed-off-by: Franz Schnyder Reviewed-by: Douglas Anderson [dianders: Adjusted Fixes/stable line based on discussion] Signed-off-by: Douglas Anderson Link: https://patch.msgid.link/20260206123758.374555-1-fra.schnyder@gmail.com --- drivers/gpu/drm/bridge/ti-sn65dsi86.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/bridge/ti-sn65dsi86.c b/drivers/gpu/drm/bridge/ti-sn65dsi86.c index 276d05d25ad8..98d64ad791d0 100644 --- a/drivers/gpu/drm/bridge/ti-sn65dsi86.c +++ b/drivers/gpu/drm/bridge/ti-sn65dsi86.c @@ -1415,6 +1415,7 @@ static int ti_sn_bridge_probe(struct auxiliary_device *adev, { struct ti_sn65dsi86 *pdata = dev_get_drvdata(adev->dev.parent); struct device_node *np = pdata->dev->of_node; + const struct i2c_client *client = to_i2c_client(pdata->dev); int ret; pdata->next_bridge = devm_drm_of_get_bridge(&adev->dev, np, 1, 0); @@ -1433,8 +1434,9 @@ static int ti_sn_bridge_probe(struct auxiliary_device *adev, ? DRM_MODE_CONNECTOR_DisplayPort : DRM_MODE_CONNECTOR_eDP; if (pdata->bridge.type == DRM_MODE_CONNECTOR_DisplayPort) { - pdata->bridge.ops = DRM_BRIDGE_OP_EDID | DRM_BRIDGE_OP_DETECT | - DRM_BRIDGE_OP_HPD; + pdata->bridge.ops = DRM_BRIDGE_OP_EDID | DRM_BRIDGE_OP_DETECT; + if (client->irq) + pdata->bridge.ops |= DRM_BRIDGE_OP_HPD; /* * If comms were already enabled they would have been enabled * with the wrong value of HPD_DISABLE. Update it now. Comms From e9e0b48cd15b46dcb2bbc165f6b0fee698b855d6 Mon Sep 17 00:00:00 2001 From: Simon Ser Date: Sun, 8 Feb 2026 22:47:26 +0000 Subject: [PATCH 06/25] drm/fourcc: fix plane order for 10/12/16-bit YCbCr formats The short comments had the correct order, but the long comments had the planes reversed. Fixes: 2271e0a20ef7 ("drm: drm_fourcc: add 10/12/16bit software decoder YCbCr formats") Signed-off-by: Simon Ser Reviewed-by: Daniel Stone Reviewed-by: Robert Mader Link: https://patch.msgid.link/20260208224718.57199-1-contact@emersion.fr --- include/uapi/drm/drm_fourcc.h | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/include/uapi/drm/drm_fourcc.h b/include/uapi/drm/drm_fourcc.h index e527b24bd824..c89aede3cb12 100644 --- a/include/uapi/drm/drm_fourcc.h +++ b/include/uapi/drm/drm_fourcc.h @@ -401,8 +401,8 @@ extern "C" { * implementation can multiply the values by 2^6=64. For that reason the padding * must only contain zeros. * index 0 = Y plane, [15:0] z:Y [6:10] little endian - * index 1 = Cr plane, [15:0] z:Cr [6:10] little endian - * index 2 = Cb plane, [15:0] z:Cb [6:10] little endian + * index 1 = Cb plane, [15:0] z:Cb [6:10] little endian + * index 2 = Cr plane, [15:0] z:Cr [6:10] little endian */ #define DRM_FORMAT_S010 fourcc_code('S', '0', '1', '0') /* 2x2 subsampled Cb (1) and Cr (2) planes 10 bits per channel */ #define DRM_FORMAT_S210 fourcc_code('S', '2', '1', '0') /* 2x1 subsampled Cb (1) and Cr (2) planes 10 bits per channel */ @@ -414,8 +414,8 @@ extern "C" { * implementation can multiply the values by 2^4=16. For that reason the padding * must only contain zeros. * index 0 = Y plane, [15:0] z:Y [4:12] little endian - * index 1 = Cr plane, [15:0] z:Cr [4:12] little endian - * index 2 = Cb plane, [15:0] z:Cb [4:12] little endian + * index 1 = Cb plane, [15:0] z:Cb [4:12] little endian + * index 2 = Cr plane, [15:0] z:Cr [4:12] little endian */ #define DRM_FORMAT_S012 fourcc_code('S', '0', '1', '2') /* 2x2 subsampled Cb (1) and Cr (2) planes 12 bits per channel */ #define DRM_FORMAT_S212 fourcc_code('S', '2', '1', '2') /* 2x1 subsampled Cb (1) and Cr (2) planes 12 bits per channel */ @@ -424,8 +424,8 @@ extern "C" { /* * 3 plane YCbCr * index 0 = Y plane, [15:0] Y little endian - * index 1 = Cr plane, [15:0] Cr little endian - * index 2 = Cb plane, [15:0] Cb little endian + * index 1 = Cb plane, [15:0] Cb little endian + * index 2 = Cr plane, [15:0] Cr little endian */ #define DRM_FORMAT_S016 fourcc_code('S', '0', '1', '6') /* 2x2 subsampled Cb (1) and Cr (2) planes 16 bits per channel */ #define DRM_FORMAT_S216 fourcc_code('S', '2', '1', '6') /* 2x1 subsampled Cb (1) and Cr (2) planes 16 bits per channel */ From 9478c166c46934160135e197b049b5a05753f2ad Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Thu, 21 Nov 2024 11:46:01 +1000 Subject: [PATCH 07/25] nouveau/gsp: drop WARN_ON in ACPI probes These WARN_ONs seem to trigger a lot, and we don't seem to have a plan to fix them, so just drop them, as they are most likely harmless. Cc: stable@vger.kernel.org Fixes: 176fdcbddfd2 ("drm/nouveau/gsp/r535: add support for booting GSP-RM") Signed-off-by: Dave Airlie Link: https://patch.msgid.link/20241121014601.229391-1-airlied@gmail.com Signed-off-by: Danilo Krummrich --- .../gpu/drm/nouveau/nvkm/subdev/gsp/rm/r535/gsp.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/rm/r535/gsp.c b/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/rm/r535/gsp.c index 7fb13434c051..a575a8dbf727 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/rm/r535/gsp.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/rm/r535/gsp.c @@ -737,8 +737,8 @@ r535_gsp_acpi_caps(acpi_handle handle, CAPS_METHOD_DATA *caps) if (!obj) goto done; - if (WARN_ON(obj->type != ACPI_TYPE_BUFFER) || - WARN_ON(obj->buffer.length != 4)) + if (obj->type != ACPI_TYPE_BUFFER || + obj->buffer.length != 4) goto done; caps->status = 0; @@ -773,8 +773,8 @@ r535_gsp_acpi_jt(acpi_handle handle, JT_METHOD_DATA *jt) if (!obj) goto done; - if (WARN_ON(obj->type != ACPI_TYPE_BUFFER) || - WARN_ON(obj->buffer.length != 4)) + if (obj->type != ACPI_TYPE_BUFFER || + obj->buffer.length != 4) goto done; jt->status = 0; @@ -861,8 +861,8 @@ r535_gsp_acpi_dod(acpi_handle handle, DOD_METHOD_DATA *dod) _DOD = output.pointer; - if (WARN_ON(_DOD->type != ACPI_TYPE_PACKAGE) || - WARN_ON(_DOD->package.count > ARRAY_SIZE(dod->acpiIdList))) + if (_DOD->type != ACPI_TYPE_PACKAGE || + _DOD->package.count > ARRAY_SIZE(dod->acpiIdList)) return; for (int i = 0; i < _DOD->package.count; i++) { From 46120745bb4e7e1f09959624716b4c5d6e2c2e9e Mon Sep 17 00:00:00 2001 From: Ethan Tidmore Date: Sun, 15 Feb 2026 22:04:38 -0600 Subject: [PATCH 08/25] drm/tiny: sharp-memory: fix pointer error dereference The function devm_drm_dev_alloc() returns a pointer error upon failure not NULL. Change null check to pointer error check. Detected by Smatch: drivers/gpu/drm/tiny/sharp-memory.c:549 sharp_memory_probe() error: 'smd' dereferencing possible ERR_PTR() Fixes: b8f9f21716fec ("drm/tiny: Add driver for Sharp Memory LCD") Signed-off-by: Ethan Tidmore Reviewed-by: Thomas Zimmermann Signed-off-by: Thomas Zimmermann Link: https://patch.msgid.link/20260216040438.43702-1-ethantidmore06@gmail.com --- drivers/gpu/drm/tiny/sharp-memory.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/tiny/sharp-memory.c b/drivers/gpu/drm/tiny/sharp-memory.c index 64272cd0f6e2..cbf69460ebf3 100644 --- a/drivers/gpu/drm/tiny/sharp-memory.c +++ b/drivers/gpu/drm/tiny/sharp-memory.c @@ -541,8 +541,8 @@ static int sharp_memory_probe(struct spi_device *spi) smd = devm_drm_dev_alloc(dev, &sharp_memory_drm_driver, struct sharp_memory_device, drm); - if (!smd) - return -ENOMEM; + if (IS_ERR(smd)) + return PTR_ERR(smd); spi_set_drvdata(spi, smd); From 7be41fb00e2c2a823f271a8318b453ca11812f1e Mon Sep 17 00:00:00 2001 From: Dan Carpenter Date: Wed, 29 Oct 2025 08:30:11 +0300 Subject: [PATCH 09/25] accel: ethosu: Fix shift overflow in cmd_to_addr() The "((cmd[0] & 0xff0000) << 16)" shift is zero. This was intended to be (((u64)cmd[0] & 0xff0000) << 16). Move the cast to the correct location. Fixes: 5a5e9c0228e6 ("accel: Add Arm Ethos-U NPU driver") Signed-off-by: Dan Carpenter Link: https://patch.msgid.link/aQGmY64tWcwOGFP4@stanley.mountain Signed-off-by: Rob Herring (Arm) --- drivers/accel/ethosu/ethosu_gem.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/accel/ethosu/ethosu_gem.c b/drivers/accel/ethosu/ethosu_gem.c index 473b5f5d7514..7b073116314b 100644 --- a/drivers/accel/ethosu/ethosu_gem.c +++ b/drivers/accel/ethosu/ethosu_gem.c @@ -154,7 +154,7 @@ static void cmd_state_init(struct cmd_state *st) static u64 cmd_to_addr(u32 *cmd) { - return ((u64)((cmd[0] & 0xff0000) << 16)) | cmd[1]; + return (((u64)cmd[0] & 0xff0000) << 16) | cmd[1]; } static u64 dma_length(struct ethosu_validated_cmdstream_info *info, From c5f8658f97ec392eeaf355d4e9775ae1f23ca1d3 Mon Sep 17 00:00:00 2001 From: Chen Ni Date: Wed, 4 Feb 2026 17:06:29 +0800 Subject: [PATCH 10/25] drm/imx: parallel-display: check return value of devm_drm_bridge_add() in imx_pd_probe() Return the value of devm_drm_bridge_add() in order to propagate the error properly, if it fails due to resource allocation failure or bridge registration failure. This ensures that the probe function fails safely rather than proceeding with a potentially incomplete bridge setup. Fixes: bf7e97910b9f ("drm/imx: parallel-display: add the bridge before attaching it") Signed-off-by: Chen Ni Reviewed-by: Luca Ceresoli Link: https://patch.msgid.link/20260204090629.2209542-1-nichen@iscas.ac.cn Signed-off-by: Luca Ceresoli --- drivers/gpu/drm/imx/ipuv3/parallel-display.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/imx/ipuv3/parallel-display.c b/drivers/gpu/drm/imx/ipuv3/parallel-display.c index 6fbf505d2801..590120a33fa0 100644 --- a/drivers/gpu/drm/imx/ipuv3/parallel-display.c +++ b/drivers/gpu/drm/imx/ipuv3/parallel-display.c @@ -256,7 +256,9 @@ static int imx_pd_probe(struct platform_device *pdev) platform_set_drvdata(pdev, imxpd); - devm_drm_bridge_add(dev, &imxpd->bridge); + ret = devm_drm_bridge_add(dev, &imxpd->bridge); + if (ret) + return ret; return component_add(dev, &imx_pd_ops); } From 496daa2759260374bb9c9b2196a849aa3bc513a8 Mon Sep 17 00:00:00 2001 From: Chen Ni Date: Fri, 6 Feb 2026 12:06:21 +0800 Subject: [PATCH 11/25] drm/bridge: synopsys: dw-dp: Check return value of devm_drm_bridge_add() in dw_dp_bind() Return the value of devm_drm_bridge_add() in order to propagate the error properly, if it fails due to resource allocation failure or bridge registration failure. This ensures that the bind function fails safely rather than proceeding with a potentially incomplete bridge setup. Fixes: b726970486d8 ("drm/bridge: synopsys: dw-dp: add bridge before attaching") Signed-off-by: Chen Ni Reviewed-by: Andy Yan Reviewed-by: Luca Ceresoli Link: https://patch.msgid.link/20260206040621.4095517-1-nichen@iscas.ac.cn Signed-off-by: Luca Ceresoli --- drivers/gpu/drm/bridge/synopsys/dw-dp.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/bridge/synopsys/dw-dp.c b/drivers/gpu/drm/bridge/synopsys/dw-dp.c index 432342452484..07f7a2e0d9f2 100644 --- a/drivers/gpu/drm/bridge/synopsys/dw-dp.c +++ b/drivers/gpu/drm/bridge/synopsys/dw-dp.c @@ -2049,7 +2049,9 @@ struct dw_dp *dw_dp_bind(struct device *dev, struct drm_encoder *encoder, bridge->type = DRM_MODE_CONNECTOR_DisplayPort; bridge->ycbcr_420_allowed = true; - devm_drm_bridge_add(dev, bridge); + ret = devm_drm_bridge_add(dev, bridge); + if (ret) + return ERR_PTR(ret); dp->aux.dev = dev; dp->aux.drm_dev = encoder->dev; From 803ec1faf7c1823e6e3b1f2aaa81be18528c9436 Mon Sep 17 00:00:00 2001 From: Osama Abdelkader Date: Mon, 9 Feb 2026 19:41:14 +0100 Subject: [PATCH 12/25] drm/bridge: samsung-dsim: Fix memory leak in error path In samsung_dsim_host_attach(), drm_bridge_add() is called to add the bridge. However, if samsung_dsim_register_te_irq() or pdata->host_ops->attach() fails afterwards, the function returns without removing the bridge, causing a memory leak. Fix this by adding proper error handling with goto labels to ensure drm_bridge_remove() is called in all error paths. Also ensure that samsung_dsim_unregister_te_irq() is called if the attach operation fails after the TE IRQ has been registered. samsung_dsim_unregister_te_irq() function is moved without changes to be before samsung_dsim_host_attach() to avoid forward declaration. Fixes: e7447128ca4a ("drm: bridge: Generalize Exynos-DSI driver into a Samsung DSIM bridge") Cc: stable@vger.kernel.org Signed-off-by: Osama Abdelkader Reviewed-by: Luca Ceresoli Link: https://patch.msgid.link/20260209184115.10937-1-osama.abdelkader@gmail.com Signed-off-by: Luca Ceresoli --- drivers/gpu/drm/bridge/samsung-dsim.c | 25 ++++++++++++++++--------- 1 file changed, 16 insertions(+), 9 deletions(-) diff --git a/drivers/gpu/drm/bridge/samsung-dsim.c b/drivers/gpu/drm/bridge/samsung-dsim.c index eabc4c32f6ab..ad8c6aa49d48 100644 --- a/drivers/gpu/drm/bridge/samsung-dsim.c +++ b/drivers/gpu/drm/bridge/samsung-dsim.c @@ -1881,6 +1881,14 @@ static int samsung_dsim_register_te_irq(struct samsung_dsim *dsi, struct device return 0; } +static void samsung_dsim_unregister_te_irq(struct samsung_dsim *dsi) +{ + if (dsi->te_gpio) { + free_irq(gpiod_to_irq(dsi->te_gpio), dsi); + gpiod_put(dsi->te_gpio); + } +} + static int samsung_dsim_host_attach(struct mipi_dsi_host *host, struct mipi_dsi_device *device) { @@ -1955,13 +1963,13 @@ static int samsung_dsim_host_attach(struct mipi_dsi_host *host, if (!(device->mode_flags & MIPI_DSI_MODE_VIDEO)) { ret = samsung_dsim_register_te_irq(dsi, &device->dev); if (ret) - return ret; + goto err_remove_bridge; } if (pdata->host_ops && pdata->host_ops->attach) { ret = pdata->host_ops->attach(dsi, device); if (ret) - return ret; + goto err_unregister_te_irq; } dsi->lanes = device->lanes; @@ -1969,14 +1977,13 @@ static int samsung_dsim_host_attach(struct mipi_dsi_host *host, dsi->mode_flags = device->mode_flags; return 0; -} -static void samsung_dsim_unregister_te_irq(struct samsung_dsim *dsi) -{ - if (dsi->te_gpio) { - free_irq(gpiod_to_irq(dsi->te_gpio), dsi); - gpiod_put(dsi->te_gpio); - } +err_unregister_te_irq: + if (!(device->mode_flags & MIPI_DSI_MODE_VIDEO)) + samsung_dsim_unregister_te_irq(dsi); +err_remove_bridge: + drm_bridge_remove(&dsi->bridge); + return ret; } static int samsung_dsim_host_detach(struct mipi_dsi_host *host, From 08fe1b5166fdc81b010d7bf39cd6440620e7931e Mon Sep 17 00:00:00 2001 From: Lizhi Hou Date: Thu, 5 Feb 2026 22:02:37 -0800 Subject: [PATCH 13/25] accel/amdxdna: Remove buffer size check when creating command BO Large command buffers may be used, and they do not always need to be mapped or accessed by the driver. Performing a size check at command BO creation time unnecessarily rejects valid use cases. Remove the buffer size check from command BO creation, and defer vmap and size validation to the paths where the driver actually needs to map and access the command buffer. Fixes: ac49797c1815 ("accel/amdxdna: Add GEM buffer object management") Reviewed-by: Mario Limonciello (AMD) Signed-off-by: Lizhi Hou Link: https://patch.msgid.link/20260206060237.4050492-1-lizhi.hou@amd.com --- drivers/accel/amdxdna/amdxdna_gem.c | 38 ++++++++++++++--------------- 1 file changed, 19 insertions(+), 19 deletions(-) diff --git a/drivers/accel/amdxdna/amdxdna_gem.c b/drivers/accel/amdxdna/amdxdna_gem.c index 8c290ddd3251..d60db49ead71 100644 --- a/drivers/accel/amdxdna/amdxdna_gem.c +++ b/drivers/accel/amdxdna/amdxdna_gem.c @@ -21,8 +21,6 @@ #include "amdxdna_pci_drv.h" #include "amdxdna_ubuf.h" -#define XDNA_MAX_CMD_BO_SIZE SZ_32K - MODULE_IMPORT_NS("DMA_BUF"); static int @@ -745,12 +743,6 @@ amdxdna_drm_create_cmd_bo(struct drm_device *dev, { struct amdxdna_dev *xdna = to_xdna_dev(dev); struct amdxdna_gem_obj *abo; - int ret; - - if (args->size > XDNA_MAX_CMD_BO_SIZE) { - XDNA_ERR(xdna, "Command bo size 0x%llx too large", args->size); - return ERR_PTR(-EINVAL); - } if (args->size < sizeof(struct amdxdna_cmd)) { XDNA_DBG(xdna, "Command BO size 0x%llx too small", args->size); @@ -764,17 +756,7 @@ amdxdna_drm_create_cmd_bo(struct drm_device *dev, abo->type = AMDXDNA_BO_CMD; abo->client = filp->driver_priv; - ret = amdxdna_gem_obj_vmap(abo, &abo->mem.kva); - if (ret) { - XDNA_ERR(xdna, "Vmap cmd bo failed, ret %d", ret); - goto release_obj; - } - return abo; - -release_obj: - drm_gem_object_put(to_gobj(abo)); - return ERR_PTR(ret); } int amdxdna_drm_create_bo_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) @@ -871,6 +853,7 @@ struct amdxdna_gem_obj *amdxdna_gem_get_obj(struct amdxdna_client *client, struct amdxdna_dev *xdna = client->xdna; struct amdxdna_gem_obj *abo; struct drm_gem_object *gobj; + int ret; gobj = drm_gem_object_lookup(client->filp, bo_hdl); if (!gobj) { @@ -879,9 +862,26 @@ struct amdxdna_gem_obj *amdxdna_gem_get_obj(struct amdxdna_client *client, } abo = to_xdna_obj(gobj); - if (bo_type == AMDXDNA_BO_INVALID || abo->type == bo_type) + if (bo_type != AMDXDNA_BO_INVALID && abo->type != bo_type) + goto put_obj; + + if (bo_type != AMDXDNA_BO_CMD || abo->mem.kva) return abo; + if (abo->mem.size > SZ_32K) { + XDNA_ERR(xdna, "Cmd bo is too big %ld", abo->mem.size); + goto put_obj; + } + + ret = amdxdna_gem_obj_vmap(abo, &abo->mem.kva); + if (ret) { + XDNA_ERR(xdna, "Vmap cmd bo failed, ret %d", ret); + goto put_obj; + } + + return abo; + +put_obj: drm_gem_object_put(gobj); return NULL; } From c68a6af400ca80596e8c37de0a1cb564aa9da8a4 Mon Sep 17 00:00:00 2001 From: Lizhi Hou Date: Thu, 5 Feb 2026 22:02:51 -0800 Subject: [PATCH 14/25] accel/amdxdna: Switch to always use chained command Preempt commands are only supported when submitted as chained commands. To ensure preempt support works consistently, always submit commands in chained command format. Set force_cmdlist to true so that single commands are filled using the chained command layout, enabling correct handling of preempt commands. Fixes: 3a0ff7b98af4 ("accel/amdxdna: Support preemption requests") Reviewed-by: Karol Wachowski Reviewed-by: Mario Limonciello (AMD) Signed-off-by: Lizhi Hou Link: https://patch.msgid.link/20260206060251.4050512-1-lizhi.hou@amd.com --- drivers/accel/amdxdna/aie2_ctx.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/accel/amdxdna/aie2_ctx.c b/drivers/accel/amdxdna/aie2_ctx.c index 4503c7c77a3e..7140c3f96362 100644 --- a/drivers/accel/amdxdna/aie2_ctx.c +++ b/drivers/accel/amdxdna/aie2_ctx.c @@ -23,9 +23,9 @@ #include "amdxdna_pci_drv.h" #include "amdxdna_pm.h" -static bool force_cmdlist; +static bool force_cmdlist = true; module_param(force_cmdlist, bool, 0600); -MODULE_PARM_DESC(force_cmdlist, "Force use command list (Default false)"); +MODULE_PARM_DESC(force_cmdlist, "Force use command list (Default true)"); #define HWCTX_MAX_TIMEOUT 60000 /* milliseconds */ From 8363c02863332992a1822688da41f881d88d1631 Mon Sep 17 00:00:00 2001 From: Lizhi Hou Date: Thu, 5 Feb 2026 22:03:06 -0800 Subject: [PATCH 15/25] accel/amdxdna: Fix crash when destroying a suspended hardware context If userspace issues an ioctl to destroy a hardware context that has already been automatically suspended, the driver may crash because the mailbox channel pointer is NULL for the suspended context. Fix this by checking the mailbox channel pointer in aie2_destroy_context() before accessing it. Fixes: 97f27573837e ("accel/amdxdna: Fix potential NULL pointer dereference in context cleanup") Reviewed-by: Karol Wachowski Signed-off-by: Lizhi Hou Link: https://patch.msgid.link/20260206060306.4050531-1-lizhi.hou@amd.com --- drivers/accel/amdxdna/aie2_message.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/drivers/accel/amdxdna/aie2_message.c b/drivers/accel/amdxdna/aie2_message.c index 7d7dcfeaf794..ab1178850c47 100644 --- a/drivers/accel/amdxdna/aie2_message.c +++ b/drivers/accel/amdxdna/aie2_message.c @@ -318,6 +318,9 @@ int aie2_destroy_context(struct amdxdna_dev_hdl *ndev, struct amdxdna_hwctx *hwc struct amdxdna_dev *xdna = ndev->xdna; int ret; + if (!hwctx->priv->mbox_chann) + return 0; + xdna_mailbox_stop_channel(hwctx->priv->mbox_chann); ret = aie2_destroy_context_req(ndev, hwctx->fw_ctx_id); xdna_mailbox_destroy_channel(hwctx->priv->mbox_chann); From 57aa3917a3b3bd805a3679371f97a1ceda3c5510 Mon Sep 17 00:00:00 2001 From: Mario Limonciello Date: Tue, 10 Feb 2026 10:42:51 -0600 Subject: [PATCH 16/25] accel/amdxdna: Reduce log noise during process termination During process termination, several error messages are logged that are not actual errors but expected conditions when a process is killed or interrupted. This creates unnecessary noise in the kernel log. The specific scenarios are: 1. HMM invalidation returns -ERESTARTSYS when the wait is interrupted by a signal during process cleanup. This is expected when a process is being terminated and should not be logged as an error. 2. Context destruction returns -ENODEV when the firmware or device has already stopped, which commonly occurs during cleanup if the device was already torn down. This is also an expected condition during orderly shutdown. Downgrade these expected error conditions from error level to debug level to reduce log noise while still keeping genuine errors visible. Fixes: 97f27573837e ("accel/amdxdna: Fix potential NULL pointer dereference in context cleanup") Reviewed-by: Lizhi Hou Signed-off-by: Mario Limonciello Signed-off-by: Lizhi Hou Link: https://patch.msgid.link/20260210164521.1094274-3-mario.limonciello@amd.com --- drivers/accel/amdxdna/aie2_ctx.c | 6 ++++-- drivers/accel/amdxdna/aie2_message.c | 4 +++- 2 files changed, 7 insertions(+), 3 deletions(-) diff --git a/drivers/accel/amdxdna/aie2_ctx.c b/drivers/accel/amdxdna/aie2_ctx.c index 7140c3f96362..e13be7608462 100644 --- a/drivers/accel/amdxdna/aie2_ctx.c +++ b/drivers/accel/amdxdna/aie2_ctx.c @@ -497,7 +497,7 @@ static void aie2_release_resource(struct amdxdna_hwctx *hwctx) if (AIE2_FEATURE_ON(xdna->dev_handle, AIE2_TEMPORAL_ONLY)) { ret = aie2_destroy_context(xdna->dev_handle, hwctx); - if (ret) + if (ret && ret != -ENODEV) XDNA_ERR(xdna, "Destroy temporal only context failed, ret %d", ret); } else { ret = xrs_release_resource(xdna->xrs_hdl, (uintptr_t)hwctx); @@ -1070,6 +1070,8 @@ void aie2_hmm_invalidate(struct amdxdna_gem_obj *abo, ret = dma_resv_wait_timeout(gobj->resv, DMA_RESV_USAGE_BOOKKEEP, true, MAX_SCHEDULE_TIMEOUT); - if (!ret || ret == -ERESTARTSYS) + if (!ret) XDNA_ERR(xdna, "Failed to wait for bo, ret %ld", ret); + else if (ret == -ERESTARTSYS) + XDNA_DBG(xdna, "Wait for bo interrupted by signal"); } diff --git a/drivers/accel/amdxdna/aie2_message.c b/drivers/accel/amdxdna/aie2_message.c index ab1178850c47..5d80c5837745 100644 --- a/drivers/accel/amdxdna/aie2_message.c +++ b/drivers/accel/amdxdna/aie2_message.c @@ -216,8 +216,10 @@ static int aie2_destroy_context_req(struct amdxdna_dev_hdl *ndev, u32 id) req.context_id = id; ret = aie2_send_mgmt_msg_wait(ndev, &msg); - if (ret) + if (ret && ret != -ENODEV) XDNA_WARN(xdna, "Destroy context failed, ret %d", ret); + else if (ret == -ENODEV) + XDNA_DBG(xdna, "Destroy context: device already stopped"); return ret; } From 1aa82181a3c285c7351523d587f7981ae4c015c8 Mon Sep 17 00:00:00 2001 From: Lizhi Hou Date: Wed, 11 Feb 2026 12:46:44 -0800 Subject: [PATCH 17/25] accel/amdxdna: Fix dead lock for suspend and resume When an application issues a query IOCTL while auto suspend is running, a deadlock can occur. The query path holds dev_lock and then calls pm_runtime_resume_and_get(), which waits for the ongoing suspend to complete. Meanwhile, the suspend callback attempts to acquire dev_lock and blocks, resulting in a deadlock. Fix this by releasing dev_lock before calling pm_runtime_resume_and_get() and reacquiring it after the call completes. Also acquire dev_lock in the resume callback to keep the locking consistent. Fixes: 063db451832b ("accel/amdxdna: Enhance runtime power management") Reviewed-by: Mario Limonciello (AMD) Signed-off-by: Lizhi Hou Link: https://patch.msgid.link/20260211204644.722758-1-lizhi.hou@amd.com --- drivers/accel/amdxdna/aie2_ctx.c | 4 ++-- drivers/accel/amdxdna/aie2_pci.c | 7 +++---- drivers/accel/amdxdna/aie2_pm.c | 2 +- drivers/accel/amdxdna/amdxdna_ctx.c | 19 +++++++------------ drivers/accel/amdxdna/amdxdna_pm.c | 2 ++ drivers/accel/amdxdna/amdxdna_pm.h | 11 +++++++++++ 6 files changed, 26 insertions(+), 19 deletions(-) diff --git a/drivers/accel/amdxdna/aie2_ctx.c b/drivers/accel/amdxdna/aie2_ctx.c index e13be7608462..8d79fafd889a 100644 --- a/drivers/accel/amdxdna/aie2_ctx.c +++ b/drivers/accel/amdxdna/aie2_ctx.c @@ -629,7 +629,7 @@ int aie2_hwctx_init(struct amdxdna_hwctx *hwctx) goto free_entity; } - ret = amdxdna_pm_resume_get(xdna); + ret = amdxdna_pm_resume_get_locked(xdna); if (ret) goto free_col_list; @@ -760,7 +760,7 @@ static int aie2_hwctx_cu_config(struct amdxdna_hwctx *hwctx, void *buf, u32 size if (!hwctx->cus) return -ENOMEM; - ret = amdxdna_pm_resume_get(xdna); + ret = amdxdna_pm_resume_get_locked(xdna); if (ret) goto free_cus; diff --git a/drivers/accel/amdxdna/aie2_pci.c b/drivers/accel/amdxdna/aie2_pci.c index 2a51b2658bfc..07e369507818 100644 --- a/drivers/accel/amdxdna/aie2_pci.c +++ b/drivers/accel/amdxdna/aie2_pci.c @@ -451,7 +451,6 @@ static int aie2_hw_suspend(struct amdxdna_dev *xdna) { struct amdxdna_client *client; - guard(mutex)(&xdna->dev_lock); list_for_each_entry(client, &xdna->client_list, node) aie2_hwctx_suspend(client); @@ -951,7 +950,7 @@ static int aie2_get_info(struct amdxdna_client *client, struct amdxdna_drm_get_i if (!drm_dev_enter(&xdna->ddev, &idx)) return -ENODEV; - ret = amdxdna_pm_resume_get(xdna); + ret = amdxdna_pm_resume_get_locked(xdna); if (ret) goto dev_exit; @@ -1044,7 +1043,7 @@ static int aie2_get_array(struct amdxdna_client *client, if (!drm_dev_enter(&xdna->ddev, &idx)) return -ENODEV; - ret = amdxdna_pm_resume_get(xdna); + ret = amdxdna_pm_resume_get_locked(xdna); if (ret) goto dev_exit; @@ -1134,7 +1133,7 @@ static int aie2_set_state(struct amdxdna_client *client, if (!drm_dev_enter(&xdna->ddev, &idx)) return -ENODEV; - ret = amdxdna_pm_resume_get(xdna); + ret = amdxdna_pm_resume_get_locked(xdna); if (ret) goto dev_exit; diff --git a/drivers/accel/amdxdna/aie2_pm.c b/drivers/accel/amdxdna/aie2_pm.c index 579b8be13b18..29bd4403a94d 100644 --- a/drivers/accel/amdxdna/aie2_pm.c +++ b/drivers/accel/amdxdna/aie2_pm.c @@ -31,7 +31,7 @@ int aie2_pm_set_dpm(struct amdxdna_dev_hdl *ndev, u32 dpm_level) { int ret; - ret = amdxdna_pm_resume_get(ndev->xdna); + ret = amdxdna_pm_resume_get_locked(ndev->xdna); if (ret) return ret; diff --git a/drivers/accel/amdxdna/amdxdna_ctx.c b/drivers/accel/amdxdna/amdxdna_ctx.c index 59fa3800b9d3..5173456bbb61 100644 --- a/drivers/accel/amdxdna/amdxdna_ctx.c +++ b/drivers/accel/amdxdna/amdxdna_ctx.c @@ -266,9 +266,9 @@ int amdxdna_drm_config_hwctx_ioctl(struct drm_device *dev, void *data, struct dr struct amdxdna_drm_config_hwctx *args = data; struct amdxdna_dev *xdna = to_xdna_dev(dev); struct amdxdna_hwctx *hwctx; - int ret, idx; u32 buf_size; void *buf; + int ret; u64 val; if (XDNA_MBZ_DBG(xdna, &args->pad, sizeof(args->pad))) @@ -310,20 +310,17 @@ int amdxdna_drm_config_hwctx_ioctl(struct drm_device *dev, void *data, struct dr return -EINVAL; } - mutex_lock(&xdna->dev_lock); - idx = srcu_read_lock(&client->hwctx_srcu); + guard(mutex)(&xdna->dev_lock); hwctx = xa_load(&client->hwctx_xa, args->handle); if (!hwctx) { XDNA_DBG(xdna, "PID %d failed to get hwctx %d", client->pid, args->handle); ret = -EINVAL; - goto unlock_srcu; + goto free_buf; } ret = xdna->dev_info->ops->hwctx_config(hwctx, args->param_type, val, buf, buf_size); -unlock_srcu: - srcu_read_unlock(&client->hwctx_srcu, idx); - mutex_unlock(&xdna->dev_lock); +free_buf: kfree(buf); return ret; } @@ -334,7 +331,7 @@ int amdxdna_hwctx_sync_debug_bo(struct amdxdna_client *client, u32 debug_bo_hdl) struct amdxdna_hwctx *hwctx; struct amdxdna_gem_obj *abo; struct drm_gem_object *gobj; - int ret, idx; + int ret; if (!xdna->dev_info->ops->hwctx_sync_debug_bo) return -EOPNOTSUPP; @@ -345,17 +342,15 @@ int amdxdna_hwctx_sync_debug_bo(struct amdxdna_client *client, u32 debug_bo_hdl) abo = to_xdna_obj(gobj); guard(mutex)(&xdna->dev_lock); - idx = srcu_read_lock(&client->hwctx_srcu); hwctx = xa_load(&client->hwctx_xa, abo->assigned_hwctx); if (!hwctx) { ret = -EINVAL; - goto unlock_srcu; + goto put_obj; } ret = xdna->dev_info->ops->hwctx_sync_debug_bo(hwctx, debug_bo_hdl); -unlock_srcu: - srcu_read_unlock(&client->hwctx_srcu, idx); +put_obj: drm_gem_object_put(gobj); return ret; } diff --git a/drivers/accel/amdxdna/amdxdna_pm.c b/drivers/accel/amdxdna/amdxdna_pm.c index d024d480521c..b1fafddd7ad5 100644 --- a/drivers/accel/amdxdna/amdxdna_pm.c +++ b/drivers/accel/amdxdna/amdxdna_pm.c @@ -16,6 +16,7 @@ int amdxdna_pm_suspend(struct device *dev) struct amdxdna_dev *xdna = to_xdna_dev(dev_get_drvdata(dev)); int ret = -EOPNOTSUPP; + guard(mutex)(&xdna->dev_lock); if (xdna->dev_info->ops->suspend) ret = xdna->dev_info->ops->suspend(xdna); @@ -28,6 +29,7 @@ int amdxdna_pm_resume(struct device *dev) struct amdxdna_dev *xdna = to_xdna_dev(dev_get_drvdata(dev)); int ret = -EOPNOTSUPP; + guard(mutex)(&xdna->dev_lock); if (xdna->dev_info->ops->resume) ret = xdna->dev_info->ops->resume(xdna); diff --git a/drivers/accel/amdxdna/amdxdna_pm.h b/drivers/accel/amdxdna/amdxdna_pm.h index 77b2d6e45570..3d26b973e0e3 100644 --- a/drivers/accel/amdxdna/amdxdna_pm.h +++ b/drivers/accel/amdxdna/amdxdna_pm.h @@ -15,4 +15,15 @@ void amdxdna_pm_suspend_put(struct amdxdna_dev *xdna); void amdxdna_pm_init(struct amdxdna_dev *xdna); void amdxdna_pm_fini(struct amdxdna_dev *xdna); +static inline int amdxdna_pm_resume_get_locked(struct amdxdna_dev *xdna) +{ + int ret; + + mutex_unlock(&xdna->dev_lock); + ret = amdxdna_pm_resume_get(xdna); + mutex_lock(&xdna->dev_lock); + + return ret; +} + #endif /* _AMDXDNA_PM_H_ */ From fdb65acfe655f844ae1e88696b9656d3ef5bb8fb Mon Sep 17 00:00:00 2001 From: Lizhi Hou Date: Wed, 11 Feb 2026 12:47:16 -0800 Subject: [PATCH 18/25] accel/amdxdna: Fix suspend failure after enabling turbo mode Enabling turbo mode disables hardware clock gating. Suspend requires hardware clock gating to be re-enabled, otherwise suspend will fail. Fix this by calling aie2_runtime_cfg() from aie2_hw_stop() to re-enable clock gating during suspend. Also ensure that firmware is initialized in aie2_hw_start() before modifying clock-gating settings during resume. Fixes: f4d7b8a6bc8c ("accel/amdxdna: Enhance power management settings") Reviewed-by: Mario Limonciello (AMD) Signed-off-by: Lizhi Hou Link: https://patch.msgid.link/20260211204716.722788-1-lizhi.hou@amd.com --- drivers/accel/amdxdna/aie2_pci.c | 13 +++++++------ 1 file changed, 7 insertions(+), 6 deletions(-) diff --git a/drivers/accel/amdxdna/aie2_pci.c b/drivers/accel/amdxdna/aie2_pci.c index 07e369507818..4b3e6bb97bd2 100644 --- a/drivers/accel/amdxdna/aie2_pci.c +++ b/drivers/accel/amdxdna/aie2_pci.c @@ -323,6 +323,7 @@ static void aie2_hw_stop(struct amdxdna_dev *xdna) return; } + aie2_runtime_cfg(ndev, AIE2_RT_CFG_CLK_GATING, NULL); aie2_mgmt_fw_fini(ndev); xdna_mailbox_stop_channel(ndev->mgmt_chann); xdna_mailbox_destroy_channel(ndev->mgmt_chann); @@ -406,18 +407,18 @@ static int aie2_hw_start(struct amdxdna_dev *xdna) goto stop_psp; } - ret = aie2_pm_init(ndev); - if (ret) { - XDNA_ERR(xdna, "failed to init pm, ret %d", ret); - goto destroy_mgmt_chann; - } - ret = aie2_mgmt_fw_init(ndev); if (ret) { XDNA_ERR(xdna, "initial mgmt firmware failed, ret %d", ret); goto destroy_mgmt_chann; } + ret = aie2_pm_init(ndev); + if (ret) { + XDNA_ERR(xdna, "failed to init pm, ret %d", ret); + goto destroy_mgmt_chann; + } + ret = aie2_mgmt_fw_query(ndev); if (ret) { XDNA_ERR(xdna, "failed to query fw, ret %d", ret); From 07efce5a6611af6714ea3ef65694e0c8dd7e44f5 Mon Sep 17 00:00:00 2001 From: Lizhi Hou Date: Wed, 11 Feb 2026 12:53:41 -0800 Subject: [PATCH 19/25] accel/amdxdna: Fix command hang on suspended hardware context When a hardware context is suspended, the job scheduler is stopped. If a command is submitted while the context is suspended, the job is queued in the scheduler but aie2_sched_job_run() is never invoked to restart the hardware context. As a result, the command hangs. Fix this by modifying the hardware context suspend routine to keep the job scheduler running so that queued jobs can trigger context restart properly. Fixes: aac243092b70 ("accel/amdxdna: Add command execution") Reviewed-by: Mario Limonciello (AMD) Signed-off-by: Lizhi Hou Link: https://patch.msgid.link/20260211205341.722982-1-lizhi.hou@amd.com --- drivers/accel/amdxdna/aie2_ctx.c | 18 +++++++++++------- 1 file changed, 11 insertions(+), 7 deletions(-) diff --git a/drivers/accel/amdxdna/aie2_ctx.c b/drivers/accel/amdxdna/aie2_ctx.c index 8d79fafd889a..25845bd5e507 100644 --- a/drivers/accel/amdxdna/aie2_ctx.c +++ b/drivers/accel/amdxdna/aie2_ctx.c @@ -53,6 +53,7 @@ static void aie2_hwctx_stop(struct amdxdna_dev *xdna, struct amdxdna_hwctx *hwct { drm_sched_stop(&hwctx->priv->sched, bad_job); aie2_destroy_context(xdna->dev_handle, hwctx); + drm_sched_start(&hwctx->priv->sched, 0); } static int aie2_hwctx_restart(struct amdxdna_dev *xdna, struct amdxdna_hwctx *hwctx) @@ -80,7 +81,6 @@ static int aie2_hwctx_restart(struct amdxdna_dev *xdna, struct amdxdna_hwctx *hw } out: - drm_sched_start(&hwctx->priv->sched, 0); XDNA_DBG(xdna, "%s restarted, ret %d", hwctx->name, ret); return ret; } @@ -297,19 +297,23 @@ aie2_sched_job_run(struct drm_sched_job *sched_job) struct dma_fence *fence; int ret; - if (!hwctx->priv->mbox_chann) + ret = amdxdna_pm_resume_get(hwctx->client->xdna); + if (ret) return NULL; - if (!mmget_not_zero(job->mm)) + if (!hwctx->priv->mbox_chann) { + amdxdna_pm_suspend_put(hwctx->client->xdna); + return NULL; + } + + if (!mmget_not_zero(job->mm)) { + amdxdna_pm_suspend_put(hwctx->client->xdna); return ERR_PTR(-ESRCH); + } kref_get(&job->refcnt); fence = dma_fence_get(job->fence); - ret = amdxdna_pm_resume_get(hwctx->client->xdna); - if (ret) - goto out; - if (job->drv_cmd) { switch (job->drv_cmd->opcode) { case SYNC_DEBUG_BO: From 1110a949675ebd56b3f0286e664ea543f745801c Mon Sep 17 00:00:00 2001 From: Lizhi Hou Date: Tue, 17 Feb 2026 10:54:15 -0800 Subject: [PATCH 20/25] accel/amdxdna: Fix out-of-bounds memset in command slot handling The remaining space in a command slot may be smaller than the size of the command header. Clearing the command header with memset() before verifying the available slot space can result in an out-of-bounds write and memory corruption. Fix this by moving the memset() call after the size validation. Fixes: 3d32eb7a5ecf ("accel/amdxdna: Fix cu_idx being cleared by memset() during command setup") Reviewed-by: Mario Limonciello (AMD) Signed-off-by: Lizhi Hou Link: https://patch.msgid.link/20260217185415.1781908-1-lizhi.hou@amd.com --- drivers/accel/amdxdna/aie2_message.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/drivers/accel/amdxdna/aie2_message.c b/drivers/accel/amdxdna/aie2_message.c index 5d80c5837745..277a27bce850 100644 --- a/drivers/accel/amdxdna/aie2_message.c +++ b/drivers/accel/amdxdna/aie2_message.c @@ -699,11 +699,11 @@ aie2_cmdlist_fill_npu_cf(struct amdxdna_gem_obj *cmd_bo, void *slot, size_t *siz u32 cmd_len; void *cmd; - memset(npu_slot, 0, sizeof(*npu_slot)); cmd = amdxdna_cmd_get_payload(cmd_bo, &cmd_len); if (*size < sizeof(*npu_slot) + cmd_len) return -EINVAL; + memset(npu_slot, 0, sizeof(*npu_slot)); npu_slot->cu_idx = amdxdna_cmd_get_cu_idx(cmd_bo); if (npu_slot->cu_idx == INVALID_CU_IDX) return -EINVAL; @@ -724,7 +724,6 @@ aie2_cmdlist_fill_npu_dpu(struct amdxdna_gem_obj *cmd_bo, void *slot, size_t *si u32 cmd_len; u32 arg_sz; - memset(npu_slot, 0, sizeof(*npu_slot)); sn = amdxdna_cmd_get_payload(cmd_bo, &cmd_len); arg_sz = cmd_len - sizeof(*sn); if (cmd_len < sizeof(*sn) || arg_sz > MAX_NPU_ARGS_SIZE) @@ -733,6 +732,7 @@ aie2_cmdlist_fill_npu_dpu(struct amdxdna_gem_obj *cmd_bo, void *slot, size_t *si if (*size < sizeof(*npu_slot) + arg_sz) return -EINVAL; + memset(npu_slot, 0, sizeof(*npu_slot)); npu_slot->cu_idx = amdxdna_cmd_get_cu_idx(cmd_bo); if (npu_slot->cu_idx == INVALID_CU_IDX) return -EINVAL; @@ -756,7 +756,6 @@ aie2_cmdlist_fill_npu_preempt(struct amdxdna_gem_obj *cmd_bo, void *slot, size_t u32 cmd_len; u32 arg_sz; - memset(npu_slot, 0, sizeof(*npu_slot)); pd = amdxdna_cmd_get_payload(cmd_bo, &cmd_len); arg_sz = cmd_len - sizeof(*pd); if (cmd_len < sizeof(*pd) || arg_sz > MAX_NPU_ARGS_SIZE) @@ -765,6 +764,7 @@ aie2_cmdlist_fill_npu_preempt(struct amdxdna_gem_obj *cmd_bo, void *slot, size_t if (*size < sizeof(*npu_slot) + arg_sz) return -EINVAL; + memset(npu_slot, 0, sizeof(*npu_slot)); npu_slot->cu_idx = amdxdna_cmd_get_cu_idx(cmd_bo); if (npu_slot->cu_idx == INVALID_CU_IDX) return -EINVAL; @@ -792,7 +792,6 @@ aie2_cmdlist_fill_npu_elf(struct amdxdna_gem_obj *cmd_bo, void *slot, size_t *si u32 cmd_len; u32 arg_sz; - memset(npu_slot, 0, sizeof(*npu_slot)); pd = amdxdna_cmd_get_payload(cmd_bo, &cmd_len); arg_sz = cmd_len - sizeof(*pd); if (cmd_len < sizeof(*pd) || arg_sz > MAX_NPU_ARGS_SIZE) @@ -801,6 +800,7 @@ aie2_cmdlist_fill_npu_elf(struct amdxdna_gem_obj *cmd_bo, void *slot, size_t *si if (*size < sizeof(*npu_slot) + arg_sz) return -EINVAL; + memset(npu_slot, 0, sizeof(*npu_slot)); npu_slot->type = EXEC_NPU_TYPE_ELF; npu_slot->inst_buf_addr = pd->inst_buf; npu_slot->save_buf_addr = pd->save_buf; From 03808abb1d868aed7478a11a82e5bb4b3f1ca6d6 Mon Sep 17 00:00:00 2001 From: Lizhi Hou Date: Tue, 17 Feb 2026 11:28:15 -0800 Subject: [PATCH 21/25] accel/amdxdna: Prevent ubuf size overflow The ubuf size calculation may overflow, resulting in an undersized allocation and possible memory corruption. Use check_add_overflow() helpers to validate the size calculation before allocation. Fixes: bd72d4acda10 ("accel/amdxdna: Support user space allocated buffer") Reviewed-by: Mario Limonciello (AMD) Signed-off-by: Lizhi Hou Link: https://patch.msgid.link/20260217192815.1784689-1-lizhi.hou@amd.com --- drivers/accel/amdxdna/amdxdna_ubuf.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) diff --git a/drivers/accel/amdxdna/amdxdna_ubuf.c b/drivers/accel/amdxdna/amdxdna_ubuf.c index b509f10b155c..fb71d6e3f44d 100644 --- a/drivers/accel/amdxdna/amdxdna_ubuf.c +++ b/drivers/accel/amdxdna/amdxdna_ubuf.c @@ -7,6 +7,7 @@ #include #include #include +#include #include #include @@ -176,7 +177,10 @@ struct dma_buf *amdxdna_get_ubuf(struct drm_device *dev, goto free_ent; } - exp_info.size += va_ent[i].len; + if (check_add_overflow(exp_info.size, va_ent[i].len, &exp_info.size)) { + ret = -EINVAL; + goto free_ent; + } } ubuf->nr_pages = exp_info.size >> PAGE_SHIFT; From 901ec3470994006bc8dd02399e16b675566c3416 Mon Sep 17 00:00:00 2001 From: Lizhi Hou Date: Thu, 19 Feb 2026 13:19:46 -0800 Subject: [PATCH 22/25] accel/amdxdna: Validate command buffer payload count The count field in the command header is used to determine the valid payload size. Verify that the valid payload does not exceed the remaining buffer space. Fixes: aac243092b70 ("accel/amdxdna: Add command execution") Reviewed-by: Mario Limonciello (AMD) Signed-off-by: Lizhi Hou Link: https://patch.msgid.link/20260219211946.1920485-1-lizhi.hou@amd.com --- drivers/accel/amdxdna/amdxdna_ctx.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/drivers/accel/amdxdna/amdxdna_ctx.c b/drivers/accel/amdxdna/amdxdna_ctx.c index 5173456bbb61..263d36072540 100644 --- a/drivers/accel/amdxdna/amdxdna_ctx.c +++ b/drivers/accel/amdxdna/amdxdna_ctx.c @@ -104,7 +104,10 @@ void *amdxdna_cmd_get_payload(struct amdxdna_gem_obj *abo, u32 *size) if (size) { count = FIELD_GET(AMDXDNA_CMD_COUNT, cmd->header); - if (unlikely(count <= num_masks)) { + if (unlikely(count <= num_masks || + count * sizeof(u32) + + offsetof(struct amdxdna_cmd, data[0]) > + abo->mem.size)) { *size = 0; return NULL; } From 74b6e83942dcc9f3cca9e561b205a5b19940a344 Mon Sep 17 00:00:00 2001 From: Matthew Brost Date: Thu, 19 Feb 2026 12:50:29 -0800 Subject: [PATCH 23/25] drm/gpusvm: Fix drm_gpusvm_pages_valid_unlocked() kernel-doc The kernel-doc for drm_gpusvm_pages_valid_unlocked() was stale and still referenced old range-based arguments and naming. Update the documentation to match the current function arguments and signature. Signed-off-by: Matthew Brost Reviewed-by: Maarten Lankhorst Link: https://patch.msgid.link/20260219205029.1011336-1-matthew.brost@intel.com --- drivers/gpu/drm/drm_gpusvm.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/drm_gpusvm.c b/drivers/gpu/drm/drm_gpusvm.c index 24180bfdf5a2..9ef9e52c0547 100644 --- a/drivers/gpu/drm/drm_gpusvm.c +++ b/drivers/gpu/drm/drm_gpusvm.c @@ -1338,14 +1338,14 @@ bool drm_gpusvm_range_pages_valid(struct drm_gpusvm *gpusvm, EXPORT_SYMBOL_GPL(drm_gpusvm_range_pages_valid); /** - * drm_gpusvm_range_pages_valid_unlocked() - GPU SVM range pages valid unlocked + * drm_gpusvm_pages_valid_unlocked() - GPU SVM pages valid unlocked * @gpusvm: Pointer to the GPU SVM structure - * @range: Pointer to the GPU SVM range structure + * @svm_pages: Pointer to the GPU SVM pages structure * - * This function determines if a GPU SVM range pages are valid. Expected be - * called without holding gpusvm->notifier_lock. + * This function determines if a GPU SVM pages are valid. Expected be called + * without holding gpusvm->notifier_lock. * - * Return: True if GPU SVM range has valid pages, False otherwise + * Return: True if GPU SVM pages are valid, False otherwise */ static bool drm_gpusvm_pages_valid_unlocked(struct drm_gpusvm *gpusvm, struct drm_gpusvm_pages *svm_pages) From c601fd5414315fc515f746b499110e46272e7243 Mon Sep 17 00:00:00 2001 From: Jonathan Cavitt Date: Tue, 24 Feb 2026 22:12:28 +0000 Subject: [PATCH 24/25] drm/client: Do not destroy NULL modes MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit 'modes' in drm_client_modeset_probe may fail to kcalloc. If this occurs, we jump to 'out', calling modes_destroy on it, which dereferences it. This may result in a NULL pointer dereference in the error case. Prevent that. Fixes: 3039cc0c0653 ("drm/client: Make copies of modes") Signed-off-by: Jonathan Cavitt Cc: Ville Syrjälä Signed-off-by: Ville Syrjälä Link: https://patch.msgid.link/20260224221227.69126-2-jonathan.cavitt@intel.com --- drivers/gpu/drm/drm_client_modeset.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/drm_client_modeset.c b/drivers/gpu/drm/drm_client_modeset.c index 262b1b8773c5..bb49b8361271 100644 --- a/drivers/gpu/drm/drm_client_modeset.c +++ b/drivers/gpu/drm/drm_client_modeset.c @@ -930,7 +930,8 @@ int drm_client_modeset_probe(struct drm_client_dev *client, unsigned int width, mutex_unlock(&client->modeset_mutex); out: kfree(crtcs); - modes_destroy(dev, modes, connector_count); + if (modes) + modes_destroy(dev, modes, connector_count); kfree(modes); kfree(offsets); kfree(enabled); From 75c151ceaacf5ca8f2f34ebf863d88002fb12587 Mon Sep 17 00:00:00 2001 From: Lizhi Hou Date: Wed, 25 Feb 2026 12:47:52 -0800 Subject: [PATCH 25/25] accel/amdxdna: Use a different name for latest firmware Using legacy driver with latest firmware causes a power off issue. Fix this by assigning a different filename (npu_7.sbin) to the latest firmware. The driver attempts to load the latest firmware first and falls back to the previous firmware version if loading fails. Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/5009 Fixes: f1eac46fe5f7 ("accel/amdxdna: Update firmware version check for latest firmware") Reviewed-by: Mario Limonciello (AMD) Signed-off-by: Lizhi Hou Link: https://patch.msgid.link/20260225204752.2711734-1-lizhi.hou@amd.com --- drivers/accel/amdxdna/aie2_pci.c | 20 +++++++++++++++++++- drivers/accel/amdxdna/amdxdna_pci_drv.c | 3 +++ drivers/accel/amdxdna/npu1_regs.c | 2 +- drivers/accel/amdxdna/npu4_regs.c | 2 +- drivers/accel/amdxdna/npu5_regs.c | 2 +- drivers/accel/amdxdna/npu6_regs.c | 2 +- 6 files changed, 26 insertions(+), 5 deletions(-) diff --git a/drivers/accel/amdxdna/aie2_pci.c b/drivers/accel/amdxdna/aie2_pci.c index 4b3e6bb97bd2..85079b6fc5d9 100644 --- a/drivers/accel/amdxdna/aie2_pci.c +++ b/drivers/accel/amdxdna/aie2_pci.c @@ -32,6 +32,11 @@ static int aie2_max_col = XRS_MAX_COL; module_param(aie2_max_col, uint, 0600); MODULE_PARM_DESC(aie2_max_col, "Maximum column could be used"); +static char *npu_fw[] = { + "npu_7.sbin", + "npu.sbin" +}; + /* * The management mailbox channel is allocated by firmware. * The related register and ring buffer information is on SRAM BAR. @@ -489,6 +494,7 @@ static int aie2_init(struct amdxdna_dev *xdna) struct psp_config psp_conf; const struct firmware *fw; unsigned long bars = 0; + char *fw_full_path; int i, nvec, ret; if (!hypervisor_is_type(X86_HYPER_NATIVE)) { @@ -503,7 +509,19 @@ static int aie2_init(struct amdxdna_dev *xdna) ndev->priv = xdna->dev_info->dev_priv; ndev->xdna = xdna; - ret = request_firmware(&fw, ndev->priv->fw_path, &pdev->dev); + for (i = 0; i < ARRAY_SIZE(npu_fw); i++) { + fw_full_path = kasprintf(GFP_KERNEL, "%s%s", ndev->priv->fw_path, npu_fw[i]); + if (!fw_full_path) + return -ENOMEM; + + ret = firmware_request_nowarn(&fw, fw_full_path, &pdev->dev); + kfree(fw_full_path); + if (!ret) { + XDNA_INFO(xdna, "Load firmware %s%s", ndev->priv->fw_path, npu_fw[i]); + break; + } + } + if (ret) { XDNA_ERR(xdna, "failed to request_firmware %s, ret %d", ndev->priv->fw_path, ret); diff --git a/drivers/accel/amdxdna/amdxdna_pci_drv.c b/drivers/accel/amdxdna/amdxdna_pci_drv.c index 4ada45d06fcf..a4384593bdcc 100644 --- a/drivers/accel/amdxdna/amdxdna_pci_drv.c +++ b/drivers/accel/amdxdna/amdxdna_pci_drv.c @@ -23,6 +23,9 @@ MODULE_FIRMWARE("amdnpu/1502_00/npu.sbin"); MODULE_FIRMWARE("amdnpu/17f0_10/npu.sbin"); MODULE_FIRMWARE("amdnpu/17f0_11/npu.sbin"); MODULE_FIRMWARE("amdnpu/17f0_20/npu.sbin"); +MODULE_FIRMWARE("amdnpu/1502_00/npu_7.sbin"); +MODULE_FIRMWARE("amdnpu/17f0_10/npu_7.sbin"); +MODULE_FIRMWARE("amdnpu/17f0_11/npu_7.sbin"); /* * 0.0: Initial version diff --git a/drivers/accel/amdxdna/npu1_regs.c b/drivers/accel/amdxdna/npu1_regs.c index 6f36a27b5a02..6e3d3ca69c04 100644 --- a/drivers/accel/amdxdna/npu1_regs.c +++ b/drivers/accel/amdxdna/npu1_regs.c @@ -72,7 +72,7 @@ static const struct aie2_fw_feature_tbl npu1_fw_feature_table[] = { }; static const struct amdxdna_dev_priv npu1_dev_priv = { - .fw_path = "amdnpu/1502_00/npu.sbin", + .fw_path = "amdnpu/1502_00/", .rt_config = npu1_default_rt_cfg, .dpm_clk_tbl = npu1_dpm_clk_table, .fw_feature_tbl = npu1_fw_feature_table, diff --git a/drivers/accel/amdxdna/npu4_regs.c b/drivers/accel/amdxdna/npu4_regs.c index a8d6f76dde5f..ce25eef5fc34 100644 --- a/drivers/accel/amdxdna/npu4_regs.c +++ b/drivers/accel/amdxdna/npu4_regs.c @@ -98,7 +98,7 @@ const struct aie2_fw_feature_tbl npu4_fw_feature_table[] = { }; static const struct amdxdna_dev_priv npu4_dev_priv = { - .fw_path = "amdnpu/17f0_10/npu.sbin", + .fw_path = "amdnpu/17f0_10/", .rt_config = npu4_default_rt_cfg, .dpm_clk_tbl = npu4_dpm_clk_table, .fw_feature_tbl = npu4_fw_feature_table, diff --git a/drivers/accel/amdxdna/npu5_regs.c b/drivers/accel/amdxdna/npu5_regs.c index c0a35cfd886c..c0ac5daf32ee 100644 --- a/drivers/accel/amdxdna/npu5_regs.c +++ b/drivers/accel/amdxdna/npu5_regs.c @@ -63,7 +63,7 @@ #define NPU5_SRAM_BAR_BASE MMNPU_APERTURE1_BASE static const struct amdxdna_dev_priv npu5_dev_priv = { - .fw_path = "amdnpu/17f0_11/npu.sbin", + .fw_path = "amdnpu/17f0_11/", .rt_config = npu4_default_rt_cfg, .dpm_clk_tbl = npu4_dpm_clk_table, .fw_feature_tbl = npu4_fw_feature_table, diff --git a/drivers/accel/amdxdna/npu6_regs.c b/drivers/accel/amdxdna/npu6_regs.c index 1fb07df99186..ce591ed0d483 100644 --- a/drivers/accel/amdxdna/npu6_regs.c +++ b/drivers/accel/amdxdna/npu6_regs.c @@ -63,7 +63,7 @@ #define NPU6_SRAM_BAR_BASE MMNPU_APERTURE1_BASE static const struct amdxdna_dev_priv npu6_dev_priv = { - .fw_path = "amdnpu/17f0_10/npu.sbin", + .fw_path = "amdnpu/17f0_10/", .rt_config = npu4_default_rt_cfg, .dpm_clk_tbl = npu4_dpm_clk_table, .fw_feature_tbl = npu4_fw_feature_table,