On Tue, Sep 05, 2023 at 09:42:28AM +0000, Maximilian Heyne wrote:
With the recent removal of vm_dev from devres its memory is only freed via the callback virtio_mmio_release_dev. However, this only takes effect after device_add is called by register_virtio_device. Until then it's an unmanaged resource and must be explicitly freed on error exit.
This bug was discovered and resolved using Coverity Static Analysis Security Testing (SAST) by Synopsys, Inc.
Cc: stable@vger.kernel.org Fixes: 55c91fedd03d ("virtio-mmio: don't break lifecycle of vm_dev") Signed-off-by: Maximilian Heyne mheyne@amazon.de
Kmemleak was reporting around 31 of these leaks (under qemu). I have not tried your patch yet as you mentioned a v2.
unreferenced object 0xffff000040bb6c00 (size 1024): comm "swapper/0", pid 1, jiffies 4294892472 (age 106.728s) hex dump (first 32 bytes): 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ backtrace: [<00000000669bb200>] __kmem_cache_alloc_node+0x1c4/0x288 [<00000000ac0129e3>] kmalloc_trace+0x20/0x2c [<0000000055d7a492>] virtio_mmio_probe+0x28/0x260 [<000000002661dcec>] platform_probe+0x68/0xdc [<00000000e9abc76a>] really_probe+0x148/0x2ac [<00000000d9fbd8d5>] __driver_probe_device+0x78/0x12c [<0000000060d7804a>] driver_probe_device+0x3c/0x15c [<00000000705fa0e9>] __driver_attach+0x94/0x19c [<000000005079c97f>] bus_for_each_dev+0x74/0xd4 [<0000000026582009>] driver_attach+0x24/0x30 [<000000003e6b0968>] bus_add_driver+0xe4/0x1e8 [<000000009e7b5554>] driver_register+0x60/0x128 [<00000000d1f180b4>] __platform_driver_register+0x28/0x34 [<0000000097a0072e>] virtio_mmio_init+0x1c/0x28 [<00000000b3340a88>] do_one_initcall+0x6c/0x1b0 [<0000000048f4ff06>] kernel_init_freeable+0x1bc/0x284
diff --git a/drivers/virtio/virtio_mmio.c b/drivers/virtio/virtio_mmio.c index 97760f611295..b2a48d07e973 100644 --- a/drivers/virtio/virtio_mmio.c +++ b/drivers/virtio/virtio_mmio.c @@ -631,13 +631,16 @@ static int virtio_mmio_probe(struct platform_device *pdev) spin_lock_init(&vm_dev->lock);
vm_dev->base = devm_platform_ioremap_resource(pdev, 0);
- if (IS_ERR(vm_dev->base))
if (IS_ERR(vm_dev->base)) {
kfree(vm_dev);return PTR_ERR(vm_dev->base);
}
/* Check magic value */ magic = readl(vm_dev->base + VIRTIO_MMIO_MAGIC_VALUE); if (magic != ('v' | 'i' << 8 | 'r' << 16 | 't' << 24)) { dev_warn(&pdev->dev, "Wrong magic value 0x%08lx!\n", magic);
kfree(vm_dev);return -ENODEV; }
@@ -646,6 +649,7 @@ static int virtio_mmio_probe(struct platform_device *pdev) if (vm_dev->version < 1 || vm_dev->version > 2) { dev_err(&pdev->dev, "Version %ld not supported!\n", vm_dev->version);
kfree(vm_dev);@@ -655,6 +659,7 @@ static int virtio_mmio_probe(struct platform_device *pdev) * virtio-mmio device with an ID 0 is a (dummy) placeholder * with no function. End probing now with no error reported. */
kfree(vm_dev);
I'd rather have a goto with a single point of freeing.
-- Catalin