Lukas Wunner wrote:
> cxl_cdat_get_length() only checks whether the DOE response size is
> sufficient for the Table Access response header (1 dword), but not the
> succeeding CDAT header (1 dword length plus other fields).
>
> It thus returns whatever uninitialized memory happens to be on the stack
> if a truncated DOE response with only 1 dword was received. Fix it.
>
> Fixes: c97006046c79 ("cxl/port: Read CDAT table")
> Reported-by: Ming Li <ming4.li@xxxxxxxxx>
> Tested-by: Ira Weiny <ira.weiny@xxxxxxxxx>
> Signed-off-by: Lukas Wunner <lukas@xxxxxxxxx>
> Cc: stable@xxxxxxxxxxxxxxx # v6.0+
> ---
> Changes v2 -> v3:
> * Newly added patch in v3
>
> drivers/cxl/core/pci.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/cxl/core/pci.c b/drivers/cxl/core/pci.c
> index d3cf1d9d67d4..11a85b3a9a0b 100644
> --- a/drivers/cxl/core/pci.c
> +++ b/drivers/cxl/core/pci.c
> @@ -528,7 +528,7 @@ static int cxl_cdat_get_length(struct device *dev,
> return rc;
> }
> wait_for_completion(&t.c);
> - if (t.task.rv < sizeof(u32))
> + if (t.task.rv < 2 * sizeof(u32))
> return -EIO;
Looks good, I wonder since this is standard for all data objects whether
the check should be pushed into the core?
For now this is easier to backport, but a follow-on could push it down a
level.
Reviewed-by: Dan Williams <dan.j.williams@xxxxxxxxx>
