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@xxxxxxxxxxxxxxx
> Fixes: 55c91fedd03d ("virtio-mmio: don't break lifecycle of vm_dev")
> Signed-off-by: Maximilian Heyne <mheyne@xxxxxxxxx>
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);
> return -ENXIO;
> }
>
> @@ -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);
> return -ENODEV;
> }
> vm_dev->vdev.id.vendor = readl(vm_dev->base + VIRTIO_MMIO_VENDOR_ID);
I'd rather have a goto with a single point of freeing.
--
Catalin
