Re: [PATCH -mm] IPS SCSI driver: Check return of scsi_add_host()

[Jeff Garzik]

Satyam Sharma wrote:
> drivers/scsi/ips.c: In function ‘ips_register_scsi’:
> drivers/scsi/ips.c:6869:
> warning: ignoring return value of ‘scsi_add_host’, declared with attribute warn_unused_result
>
> scsi_add_host() is __must_check, so let's check it's return and cleanup
> appropriately on errors.
>
> Signed-off-by: Satyam Sharma <satyam@xxxxxxxxxxxxx>
>
> ---
>
>  drivers/scsi/ips.c |    7 ++++++-
>  1 file changed, 6 insertions(+), 1 deletion(-)
>
> --- linux-2.6.23-rc4-mm1/drivers/scsi/ips.c~fix	2007-09-02 20:21:27.000000000 +0530
> +++ linux-2.6.23-rc4-mm1/drivers/scsi/ips.c	2007-09-02 20:25:58.000000000 +0530
> @@ -6866,7 +6866,12 @@ ips_register_scsi(int index)
>  	sh->max_channel = ha->nbus - 1;
>  	sh->can_queue = ha->max_cmds - 1;
>  
> -	scsi_add_host(sh, NULL);
> +	if (scsi_add_host(sh, NULL)) {
> +		IPS_PRINTK(KERN_WARNING, ha->pcidev, "Unable to add SCSI host\n");
> +		free_irq(ha->irq, ha);
> +		scsi_host_put(sh);
> +		return -1;

ACK, as long as you add a comment to this code and the request_irq() 
failure code:  the oldha/old irq is not restored upon failure, in this 
function.

That is a pre-existing bug, and not your fault, but your patch 
"continues in the same buggy tradition" :)  We should at least note the 
FIXME at each error handling code branch.

Ideally you or somebody should do a detailed analysis to
a) (preferably) get rid of the silly oldha/ double-irq-request weirdness 
and make it look like other drivers,
	or,
b) analyze the code and see what it takes to _really_ unwind the error.

Also, I would recommend moving the error handling code to the end of the 
function and using the standard 'goto' approach for function error 
handling.  This eliminates the duplicate scsi_host_put() calls upon error.

	Jeff





-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/