From: Markus Elfring <[email protected]>
Date: Tue, 27 Feb 2024 14:05:25 +0100
Adjust jump targets so that a bit of exception handling can be better
reused at the end of this function implementation.
Signed-off-by: Markus Elfring <[email protected]>
---
.../net/ethernet/freescale/fman/fman_dtsec.c | 19 +++++++++++--------
1 file changed, 11 insertions(+), 8 deletions(-)
diff --git a/drivers/net/ethernet/freescale/fman/fman_dtsec.c b/drivers/net/ethernet/freescale/fman/fman_dtsec.c
index 3088da7adf0f..1de22400fd89 100644
--- a/drivers/net/ethernet/freescale/fman/fman_dtsec.c
+++ b/drivers/net/ethernet/freescale/fman/fman_dtsec.c
@@ -1279,9 +1279,8 @@ static int dtsec_init(struct fman_mac *dtsec)
dtsec->max_speed, dtsec->addr, dtsec->exceptions,
dtsec->tbidev->addr);
if (err) {
- free_init_resources(dtsec);
pr_err("DTSEC version doesn't support this i/f mode\n");
- return err;
+ goto free_resources;
}
/* Configure the TBI PHY Control Register */
@@ -1296,23 +1295,21 @@ static int dtsec_init(struct fman_mac *dtsec)
err = fman_set_mac_max_frame(dtsec->fm, dtsec->mac_id, max_frm_ln);
if (err) {
pr_err("Setting max frame length failed\n");
- free_init_resources(dtsec);
- return -EINVAL;
+ err = -EINVAL;
+ goto free_resources;
}
dtsec->multicast_addr_hash =
alloc_hash_table(EXTENDED_HASH_TABLE_SIZE);
if (!dtsec->multicast_addr_hash) {
- free_init_resources(dtsec);
pr_err("MC hash table is failed\n");
- return -ENOMEM;
+ goto e_nomem;
}
dtsec->unicast_addr_hash = alloc_hash_table(DTSEC_HASH_TABLE_SIZE);
if (!dtsec->unicast_addr_hash) {
- free_init_resources(dtsec);
pr_err("UC hash table is failed\n");
- return -ENOMEM;
+ goto e_nomem;
}
/* register err intr handler for dtsec to FPM (err) */
@@ -1326,6 +1323,12 @@ static int dtsec_init(struct fman_mac *dtsec)
dtsec->dtsec_drv_param = NULL;
return 0;
+
+e_nomem:
+ err = -ENOMEM;
+free_resources:
+ free_init_resources(dtsec);
+ return err;
}
static int dtsec_free(struct fman_mac *dtsec)
--
2.43.2
> -----Original Message-----
> From: Markus Elfring <[email protected]>
> Sent: Tuesday, February 27, 2024 3:15 PM
> To: [email protected]; [email protected]; David S. Miller
> <[email protected]>; Eric Dumazet <[email protected]>; Jakub
> Kicinski <[email protected]>; Madalin Bucur <[email protected]>; Paolo
> Abeni <[email protected]>; Sean Anderson <[email protected]>
> Cc: LKML <[email protected]>
> Subject: [PATCH] net: fman: Use common error handling code in dtsec_init()
>
> From: Markus Elfring <[email protected]>
> Date: Tue, 27 Feb 2024 14:05:25 +0100
>
> Adjust jump targets so that a bit of exception handling can be better
> reused at the end of this function implementation.
>
> Signed-off-by: Markus Elfring <[email protected]>
> ---
> .../net/ethernet/freescale/fman/fman_dtsec.c | 19 +++++++++++--------
> 1 file changed, 11 insertions(+), 8 deletions(-)
>
> diff --git a/drivers/net/ethernet/freescale/fman/fman_dtsec.c
> b/drivers/net/ethernet/freescale/fman/fman_dtsec.c
> index 3088da7adf0f..1de22400fd89 100644
> --- a/drivers/net/ethernet/freescale/fman/fman_dtsec.c
> +++ b/drivers/net/ethernet/freescale/fman/fman_dtsec.c
> @@ -1279,9 +1279,8 @@ static int dtsec_init(struct fman_mac *dtsec)
> dtsec->max_speed, dtsec->addr, dtsec->exceptions,
> dtsec->tbidev->addr);
> if (err) {
> - free_init_resources(dtsec);
> pr_err("DTSEC version doesn't support this i/f mode\n");
> - return err;
> + goto free_resources;
> }
>
> /* Configure the TBI PHY Control Register */
> @@ -1296,23 +1295,21 @@ static int dtsec_init(struct fman_mac *dtsec)
> err = fman_set_mac_max_frame(dtsec->fm, dtsec->mac_id,
> max_frm_ln);
> if (err) {
> pr_err("Setting max frame length failed\n");
> - free_init_resources(dtsec);
> - return -EINVAL;
> + err = -EINVAL;
> + goto free_resources;
> }
>
> dtsec->multicast_addr_hash =
> alloc_hash_table(EXTENDED_HASH_TABLE_SIZE);
> if (!dtsec->multicast_addr_hash) {
> - free_init_resources(dtsec);
> pr_err("MC hash table is failed\n");
> - return -ENOMEM;
> + goto e_nomem;
> }
>
> dtsec->unicast_addr_hash =
> alloc_hash_table(DTSEC_HASH_TABLE_SIZE);
> if (!dtsec->unicast_addr_hash) {
> - free_init_resources(dtsec);
> pr_err("UC hash table is failed\n");
> - return -ENOMEM;
> + goto e_nomem;
> }
>
> /* register err intr handler for dtsec to FPM (err) */
> @@ -1326,6 +1323,12 @@ static int dtsec_init(struct fman_mac *dtsec)
> dtsec->dtsec_drv_param = NULL;
>
> return 0;
> +
> +e_nomem:
> + err = -ENOMEM;
> +free_resources:
> + free_init_resources(dtsec);
> + return err;
> }
>
> static int dtsec_free(struct fman_mac *dtsec)
> --
> 2.43.2
Looks good
Acked-by: Madalin Bucur <[email protected]>
Tue, Feb 27, 2024 at 02:14:52PM CET, [email protected] wrote:
>From: Markus Elfring <[email protected]>
>Date: Tue, 27 Feb 2024 14:05:25 +0100
>
>Adjust jump targets so that a bit of exception handling can be better
>reused at the end of this function implementation.
>
>Signed-off-by: Markus Elfring <[email protected]>
Reviewed-by: Jiri Pirko <[email protected]>
Nit, next time please indicate the target tree in the patch subject
prefix: [patch net-next] xxx
On Tue, 27 Feb 2024 14:14:52 +0100 Markus Elfring wrote:
> Adjust jump targets so that a bit of exception handling can be better
> reused at the end of this function implementation.
Okay, but..
> .../net/ethernet/freescale/fman/fman_dtsec.c | 19 +++++++++++--------
> 1 file changed, 11 insertions(+), 8 deletions(-)
.you've added more lines than you've removed so what's the point.
--
pw-bot: reject
Wed, Feb 28, 2024 at 03:46:57AM CET, [email protected] wrote:
>On Tue, 27 Feb 2024 14:14:52 +0100 Markus Elfring wrote:
>> Adjust jump targets so that a bit of exception handling can be better
>> reused at the end of this function implementation.
>
>Okay, but..
>
>> .../net/ethernet/freescale/fman/fman_dtsec.c | 19 +++++++++++--------
>> 1 file changed, 11 insertions(+), 8 deletions(-)
>
>..you've added more lines than you've removed so what's the point.
To have cleaner error path? Not always lines of code is the correct
indicator of patch quality :)
>--
>pw-bot: reject
>
>> Adjust jump targets so that a bit of exception handling can be better
>> reused at the end of this function implementation.
>
> Okay, but..
>
>> .../net/ethernet/freescale/fman/fman_dtsec.c | 19 +++++++++++--------
>> 1 file changed, 11 insertions(+), 8 deletions(-)
>
> ..you've added more lines than you've removed so what's the point.
* Can the change acceptance grow any more for attempts to improve error handling
another bit?
* Would you like to fiddle with scope-based resource management?
https://elixir.bootlin.com/linux/v6.8-rc6/source/include/linux/cleanup.h#L8
See also:
Article by Jonathan Corbet from 2023-06-15
https://lwn.net/Articles/934679/
> --
> pw-bot: reject
Can such an reaction be reconsidered once more?
Regards,
Markus
On Wed, Feb 28, 2024 at 08:51:49AM +0100, Jiri Pirko wrote:
> Wed, Feb 28, 2024 at 03:46:57AM CET, [email protected] wrote:
> >On Tue, 27 Feb 2024 14:14:52 +0100 Markus Elfring wrote:
> >> Adjust jump targets so that a bit of exception handling can be better
> >> reused at the end of this function implementation.
> >
> >Okay, but..
> >
> >> .../net/ethernet/freescale/fman/fman_dtsec.c | 19 +++++++++++--------
> >> 1 file changed, 11 insertions(+), 8 deletions(-)
> >
> >..you've added more lines than you've removed so what's the point.
>
> To have cleaner error path? Not always lines of code is the correct
> indicator of patch quality :)
>
I really don't like those goto e_nomem type of things. When you're
laddering gotos you should do that kind of thing before the gotos so
that when people add new gotos it doesn't make a mess. It's the same
for unlocks, do that before the goto unless it matches a lock at the
very start of the function. Or if you're doing a goto from inside a
loop then clean up the partial iteration through the loop before the
goto.
regards,
dan carpenter