On 11/8/22 16:52, Yang Yingliang wrote:
> In ata_tdev_add(), the return value of transport_add_device() is not
> checked. As a result, another error after that function call leads to
> a kernel crash (null-ptr-deref) because transport_remove_device() is
> called to remove a device that was not added.
This was the error pattern for the previous 2 cases, but not for this
one. For this case, transport_configure_device() would be called for a
device that was not added, no ? so where does the backtrace below come
from ? It does not correspond to the code this patch is changing.
>
> Unable to handle kernel NULL pointer dereference at virtual address 00000000000000d0
> CPU: 61 PID: 13969 Comm: rmmod Kdump: loaded Tainted: G W 6.1.0-rc3+ #13
> pstate: 60400009 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
> pc : device_del+0x48/0x39c
> lr : device_del+0x44/0x39c
> Call trace:
> device_del+0x48/0x39c
> attribute_container_class_device_del+0x28/0x40
> transport_remove_classdev+0x60/0x7c
> attribute_container_device_trigger+0x118/0x120
> transport_remove_device+0x20/0x30
> ata_tlink_delete+0x4c/0xb0 [libata]
> ata_tport_delete+0x2c/0x60 [libata]
> ata_port_detach+0x148/0x1b0 [libata]
> ata_pci_remove_one+0x50/0x80 [libata]
> ahci_remove_one+0x4c/0x8c [ahci]
>
> Fix this by checking and handling return value of transport_add_device()
> in ata_tdev_add().
>
> Fixes: d9027470b886 ("[libata] Add ATA transport class")
> Signed-off-by: Yang Yingliang <yangyingliang@xxxxxxxxxx>
> ---
> drivers/ata/libata-transport.c | 8 +++++++-
> 1 file changed, 7 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/ata/libata-transport.c b/drivers/ata/libata-transport.c
> index aac9336e8153..e4fb9d1b9b39 100644
> --- a/drivers/ata/libata-transport.c
> +++ b/drivers/ata/libata-transport.c
> @@ -713,7 +713,13 @@ static int ata_tdev_add(struct ata_device *ata_dev)
> return error;
> }
>
> - transport_add_device(dev);
> + error = transport_add_device(dev);
> + if (error) {
> + device_del(dev);
> + ata_tdev_free(ata_dev);
> + return error;
> + }
> +
> transport_configure_device(dev);
> return 0;
> }
--
Damien Le Moal
Western Digital Research
