AMD General

Reviewed-by: YiPeng Chai <[email protected]>

Best Regards,
Thomas
________________________________
From: Xie, Chenglei <[email protected]>
Sent: Tuesday, May 12, 2026 3:29 AM
To: [email protected] <[email protected]>
Cc: Chan, Hing Pong <[email protected]>; Luo, Zhigang <[email protected]>; 
Deucher, Alexander <[email protected]>; Xie, Chenglei 
<[email protected]>; Chai, Thomas <[email protected]>
Subject: [PATCH] drm/amdgpu: bound SR-IOV RAS CPER dump parsing against 
used_size

The VF copies a PF-provided CPER telemetry blob and walks records using
cper_dump->count and each entry's record_length. count is u64 while the
loop used u32, so a large count could loop indefinitely. record_length was
not limited to the kmemdup'd region, so the first iteration could read far
past the allocation; record_length == 0 could spin forever on the same
entry. Together that allowed a malicious hypervisor to leak heap past the
blob into the CPER ring or hang the guest.

Require used_size to cover the fixed header before buf and stay within the
telemetry cap. Track remaining bytes in buf, cap iterations with u64 and
CPER_MAX_ALLOWED_COUNT, and reject record_length outside
[sizeof(cper_hdr), remaining] before writing to the ring.

Signed-off-by: Chenglei Xie <[email protected]>
Change-Id: Ic21f4523eebc6c4b4f8c6b62b84104b18cf86a48
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c | 20 +++++++++++++++-----
 1 file changed, 15 insertions(+), 5 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c 
b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
index 6974b1c5b56c2..c8bec62bdffb2 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
@@ -1798,13 +1798,15 @@ amdgpu_virt_write_cpers_to_ring(struct amdgpu_device 
*adev,
         struct amd_sriov_ras_cper_dump *cper_dump = NULL;
         struct cper_hdr *entry = NULL;
         struct amdgpu_ring *ring = &adev->cper.ring_buf;
-       uint32_t checksum, used_size, i;
+       uint32_t checksum, used_size;
+       u64 remaining, cnt, i;
         int ret = 0;

         checksum = host_telemetry->header.checksum;
         used_size = host_telemetry->header.used_size;

-       if (used_size > (AMD_SRIOV_MSG_RAS_TELEMETRY_SIZE_KB_V1 << 10))
+       if (used_size < offsetof(struct amd_sriov_ras_cper_dump, buf) ||
+           used_size > (AMD_SRIOV_MSG_RAS_TELEMETRY_SIZE_KB_V1 << 10))
                 return -EINVAL;

         cper_dump = kmemdup(&host_telemetry->body.cper_dump, used_size, 
GFP_KERNEL);
@@ -1829,11 +1831,19 @@ amdgpu_virt_write_cpers_to_ring(struct amdgpu_device 
*adev,
         }

         entry = (struct cper_hdr *)&cper_dump->buf[0];
+       remaining = (u64)used_size - offsetof(struct amd_sriov_ras_cper_dump, 
buf);
+       cnt = min_t(u64, cper_dump->count, CPER_MAX_ALLOWED_COUNT);
+
+       for (i = 0; i < cnt; i++) {
+               if (entry->record_length < sizeof(struct cper_hdr) ||
+                   entry->record_length > remaining) {
+                       ret = -EINVAL;
+                       goto out;
+               }

-       for (i = 0; i < cper_dump->count; i++) {
                 amdgpu_cper_ring_write(ring, entry, entry->record_length);
-               entry = (struct cper_hdr *)((char *)entry +
-                                           entry->record_length);
+               remaining -= entry->record_length;
+               entry = (struct cper_hdr *)((char *)entry + 
entry->record_length);
         }

         if (cper_dump->overflow_count)
--
2.34.1

Reply via email to