diff --git a/src/dawn/common/GPUInfo.cpp b/src/dawn/common/GPUInfo.cpp
index 97068ff6ca..ddd8459703 100644
--- a/src/dawn/common/GPUInfo.cpp
+++ b/src/dawn/common/GPUInfo.cpp
@@ -63,6 +63,9 @@ namespace gpu_info {
     bool IsIntel(PCIVendorID vendorId) {
         return vendorId == kVendorID_Intel;
     }
+    bool IsMesa(PCIVendorID vendorId) {
+        return vendorId == kVendorID_Mesa;
+    }
     bool IsNvidia(PCIVendorID vendorId) {
         return vendorId == kVendorID_Nvidia;
     }
diff --git a/src/dawn/common/GPUInfo.h b/src/dawn/common/GPUInfo.h
index 2b30d96662..26c91036e6 100644
--- a/src/dawn/common/GPUInfo.h
+++ b/src/dawn/common/GPUInfo.h
@@ -27,6 +27,7 @@ namespace gpu_info {
     static constexpr PCIVendorID kVendorID_ARM = 0x13B5;
     static constexpr PCIVendorID kVendorID_ImgTec = 0x1010;
     static constexpr PCIVendorID kVendorID_Intel = 0x8086;
+    static constexpr PCIVendorID kVendorID_Mesa = 0x10005;
     static constexpr PCIVendorID kVendorID_Nvidia = 0x10DE;
     static constexpr PCIVendorID kVendorID_Qualcomm = 0x5143;
     static constexpr PCIVendorID kVendorID_Google = 0x1AE0;
@@ -39,6 +40,7 @@ namespace gpu_info {
     bool IsARM(PCIVendorID vendorId);
     bool IsImgTec(PCIVendorID vendorId);
     bool IsIntel(PCIVendorID vendorId);
+    bool IsMesa(PCIVendorID vendorId);
     bool IsNvidia(PCIVendorID vendorId);
     bool IsQualcomm(PCIVendorID vendorId);
     bool IsSwiftshader(PCIVendorID vendorId, PCIDeviceID deviceId);
diff --git a/src/dawn/native/vulkan/AdapterVk.cpp b/src/dawn/native/vulkan/AdapterVk.cpp
index 6957a39406..4437882295 100644
--- a/src/dawn/native/vulkan/AdapterVk.cpp
+++ b/src/dawn/native/vulkan/AdapterVk.cpp
@@ -285,10 +285,10 @@ namespace dawn::native::vulkan {
 
         // Only check maxFragmentCombinedOutputResources on mobile GPUs. Desktop GPUs drivers seem
         // to put incorrect values for this limit with things like 8 or 16 when they can do bindless
-        // storage buffers.
+        // storage buffers. Mesa llvmpipe driver also puts 8 here.
         uint32_t vendorId = mDeviceInfo.properties.vendorID;
         if (!gpu_info::IsAMD(vendorId) && !gpu_info::IsIntel(vendorId) &&
-            !gpu_info::IsNvidia(vendorId)) {
+            !gpu_info::IsMesa(vendorId) && !gpu_info::IsNvidia(vendorId)) {
             if (vkLimits.maxFragmentCombinedOutputResources <
                 kMaxColorAttachments + baseLimits.v1.maxStorageTexturesPerShaderStage +
                     baseLimits.v1.maxStorageBuffersPerShaderStage) {