search for: pmemory

When resuming from ACPI S3, memory leak happens in nvkm_ram_get(). This is because *pmemory points to newly allocated memory without checking and freeing the old memory. Here is the log showing this issue. unreferenced object 0xffffa3b608c6d5c0 (size 64): comm "kworker/u32:30", pid 934, jiffies 4294823520 (age 5000.217s) hex dump (first 32 bytes): 00 fc 4a c0 ff ff ff...

...s = &gk20a_instobj_ptrs; /* Allocate backing memory */ for (i = 0; i < npages; i++) { @@ -533,7 +533,7 @@ gk20a_instobj_new(struct nvkm_instmem *base, u32 size, u32 align, bool zero, else ret = gk20a_instobj_ctor_dma(imem, size >> PAGE_SHIFT, align, &node); - *pmemory = node ? &node->memory : NULL; + *pmemory = node ? &node->base.memory : NULL; if (ret) return ret; -- 2.43.0

..., 2015, leads to the following Smatch complaint: drivers/gpu/drm/nouveau/nvkm/subdev/instmem/gk20a.c:536 gk20a_instobj_new() error: we previously assumed 'node' could be null (see line 532) drivers/gpu/drm/nouveau/nvkm/subdev/instmem/gk20a.c 531 align, &node); 532 *pmemory = node ? &node->memory : NULL; ^^^^ Added a new check for NULL. 533 if (ret) 534 return ret; 535 536 node->imem = imem; ^^^^^^^^^^ Unchecked dereference. 537 538 /* present memory for being mapped using small pages */...

...* Allocate backing memory */ > for (i = 0; i < npages; i++) { > @@ -533,7 +533,7 @@ gk20a_instobj_new(struct nvkm_instmem *base, u32 size, u32 align, bool zero, > else > ret = gk20a_instobj_ctor_dma(imem, size >> PAGE_SHIFT, > align, &node); > - *pmemory = node ? &node->memory : NULL; > + *pmemory = node ? &node->base.memory : NULL; > if (ret) > return ret; > Tested-by: Jon Hunter <jonathanh at nvidia.com> Thanks! Jon -- nvpublic