From 6d2191d226553c86f8ea98f5e6cba13d9ef13280 Mon Sep 17 00:00:00 2001 From: Ellen Pan Date: Tue, 7 Oct 2025 09:46:16 -0500 Subject: drm/amdgpu: Add SRIOV crit_region_version support 1. Added enum amd_sriov_crit_region_version to support multi versions 2. Added logic in SRIOV mailbox to regonize crit_region version during req_gpu_init_data Signed-off-by: Ellen Pan Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h index d1172c8e58c4..36247a160aa6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h @@ -262,6 +262,11 @@ struct amdgpu_virt_ras { DECLARE_ATTR_CAP_CLASS(amdgpu_virt, AMDGPU_VIRT_CAPS_LIST); +struct amdgpu_virt_region { + uint32_t offset; + uint32_t size_kb; +}; + /* GPU virtualization */ struct amdgpu_virt { uint32_t caps; @@ -289,6 +294,9 @@ struct amdgpu_virt { bool ras_init_done; uint32_t reg_access; + /* dynamic(v2) critical regions */ + struct amdgpu_virt_region init_data_header; + /* vf2pf message */ struct delayed_work vf2pf_work; uint32_t vf2pf_update_interval_ms; -- cgit v1.2.3 From 07009df6494dc9272b233eeddc10d26f545ff09a Mon Sep 17 00:00:00 2001 From: Ellen Pan Date: Tue, 7 Oct 2025 11:00:16 -0500 Subject: drm/amdgpu: Introduce SRIOV critical regions v2 during VF init 1. Introduced amdgpu_virt_init_critical_region during VF init. - VFs use init_data_header_offset and init_data_header_size_kb transmitted via PF2VF mailbox to fetch the offset of critical regions' offsets/sizes in VRAM and save to adev->virt.crit_region_offsets and adev->virt.crit_region_sizes_kb. Signed-off-by: Ellen Pan Reviewed-by: Alex Deucher Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h | 11 +++++++++++ 1 file changed, 11 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h index 36247a160aa6..8d03a8620de9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h @@ -54,6 +54,12 @@ #define AMDGPU_VF2PF_UPDATE_MAX_RETRY_LIMIT 2 +/* Signature used to validate the SR-IOV dynamic critical region init data header ("INDA") */ +#define AMDGPU_SRIOV_CRIT_DATA_SIGNATURE "INDA" +#define AMDGPU_SRIOV_CRIT_DATA_SIG_LEN 4 + +#define IS_SRIOV_CRIT_REGN_ENTRY_VALID(hdr, id) ((hdr)->valid_tables & (1 << (id))) + enum amdgpu_sriov_vf_mode { SRIOV_VF_MODE_BARE_METAL = 0, SRIOV_VF_MODE_ONE_VF, @@ -296,6 +302,9 @@ struct amdgpu_virt { /* dynamic(v2) critical regions */ struct amdgpu_virt_region init_data_header; + struct amdgpu_virt_region crit_regn; + struct amdgpu_virt_region crit_regn_tbl[AMD_SRIOV_MSG_MAX_TABLE_ID]; + bool is_dynamic_crit_regn_enabled; /* vf2pf message */ struct delayed_work vf2pf_work; @@ -432,6 +441,8 @@ void amdgpu_virt_exchange_data(struct amdgpu_device *adev); void amdgpu_virt_fini_data_exchange(struct amdgpu_device *adev); void amdgpu_virt_init(struct amdgpu_device *adev); +int amdgpu_virt_init_critical_region(struct amdgpu_device *adev); + bool amdgpu_virt_can_access_debugfs(struct amdgpu_device *adev); int amdgpu_virt_enable_access_debugfs(struct amdgpu_device *adev); void amdgpu_virt_disable_access_debugfs(struct amdgpu_device *adev); -- cgit v1.2.3 From b4a8fcc7826ad2bc95f05c8481d122ba23535a65 Mon Sep 17 00:00:00 2001 From: Ellen Pan Date: Tue, 7 Oct 2025 11:12:39 -0500 Subject: drm/amdgpu: Add logic for VF ipd and VF bios to init from dynamic crit_region offsets 1. Added VF logic in amdgpu_virt to init IP discovery using the offsets from dynamic(v2) critical regions; 2. Added VF logic in amdgpu_virt to init bios image using the offsets from dynamic(v2) critical regions; Signed-off-by: Ellen Pan Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h index 8d03a8620de9..2a13cc892a13 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h @@ -442,6 +442,8 @@ void amdgpu_virt_fini_data_exchange(struct amdgpu_device *adev); void amdgpu_virt_init(struct amdgpu_device *adev); int amdgpu_virt_init_critical_region(struct amdgpu_device *adev); +int amdgpu_virt_get_dynamic_data_info(struct amdgpu_device *adev, + int data_id, uint8_t *binary, uint64_t *size); bool amdgpu_virt_can_access_debugfs(struct amdgpu_device *adev); int amdgpu_virt_enable_access_debugfs(struct amdgpu_device *adev); -- cgit v1.2.3 From 90ef1dcb1d2bb84ad998e845e26a2a297a7ddfd6 Mon Sep 17 00:00:00 2001 From: Srinivasan Shanmugam Date: Thu, 23 Oct 2025 10:52:21 +0530 Subject: drm/amdgpu: Fix pointer casts when reading dynamic region sizes MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The function amdgpu_virt_get_dynamic_data_info() writes a 64-bit size value. In two places (amdgpu_bios.c and amdgpu_discovery.c), the code passed the address of a smaller variable by casting it to u64 *, which is unsafe. This could make the function write more bytes than the smaller variable can hold, possibly overwriting nearby memory. Reported by static analysis tools. v2: Dynamic region size comes from the host (SR-IOV setup) and is always fixed to 5 MB. (Lijo/Ellen) 5 MB easily fits inside a 32-bit value, so using a 64-bit type is not needed. It also avoids extra type casts Fixes: b4a8fcc7826a ("drm/amdgpu: Add logic for VF ipd and VF bios to init from dynamic crit_region offsets") Reported by: Dan Carpenter Cc: Ellen Pan Cc: Christian König Cc: Alex Deucher Signed-off-by: Srinivasan Shanmugam Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h index 2a13cc892a13..14d864be5800 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h @@ -443,7 +443,7 @@ void amdgpu_virt_init(struct amdgpu_device *adev); int amdgpu_virt_init_critical_region(struct amdgpu_device *adev); int amdgpu_virt_get_dynamic_data_info(struct amdgpu_device *adev, - int data_id, uint8_t *binary, uint64_t *size); + int data_id, uint8_t *binary, u32 *size); bool amdgpu_virt_can_access_debugfs(struct amdgpu_device *adev); int amdgpu_virt_enable_access_debugfs(struct amdgpu_device *adev); -- cgit v1.2.3 From 4d0e55965c9c2f085a83f6a69acd7b5a44132de5 Mon Sep 17 00:00:00 2001 From: Will Aitken Date: Tue, 7 Oct 2025 14:19:45 +0000 Subject: drm/amdgpu: Update headers for sriov xgmi ext peer link support feature flag Adds new sriov msg flag to match host, feature flag in the amdgim enum, and a wrapper macro to check it. Signed-off-by: Will Aitken Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h index 14d864be5800..4fd194a9a972 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h @@ -150,6 +150,7 @@ enum AMDGIM_FEATURE_FLAG { AMDGIM_FEATURE_RAS_CAPS = (1 << 9), AMDGIM_FEATURE_RAS_TELEMETRY = (1 << 10), AMDGIM_FEATURE_RAS_CPER = (1 << 11), + AMDGIM_FEATURE_XGMI_TA_EXT_PEER_LINK = (1 << 12), }; enum AMDGIM_REG_ACCESS_FLAG { @@ -395,6 +396,9 @@ struct amdgpu_video_codec_info; #define amdgpu_sriov_ras_cper_en(adev) \ ((adev)->virt.gim_feature & AMDGIM_FEATURE_RAS_CPER) +#define amdgpu_sriov_xgmi_ta_ext_peer_link_en(adev) \ +((adev)->virt.gim_feature & AMDGIM_FEATURE_XGMI_TA_EXT_PEER_LINK) + static inline bool is_virtual_machine(void) { #if defined(CONFIG_X86) -- cgit v1.2.3 From 2b198d459f0983313fc090b1527c68569f406575 Mon Sep 17 00:00:00 2001 From: YiPeng Chai Date: Mon, 21 Jul 2025 15:22:27 +0800 Subject: drm/amdgpu: Add lock to serialize sriov command execution Add lock to serialize sriov command execution. Signed-off-by: YiPeng Chai Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h index 4fd194a9a972..01d5bca2dee1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h @@ -325,6 +325,8 @@ struct amdgpu_virt { /* Spinlock to protect access to the RLCG register interface */ spinlock_t rlcg_reg_lock; + struct mutex access_req_mutex; + union amd_sriov_ras_caps ras_en_caps; union amd_sriov_ras_caps ras_telemetry_en_caps; struct amdgpu_virt_ras ras; -- cgit v1.2.3