From: Jack MIN <jackmin@mellanox.com>
To: Yongseok Koh <yskoh@mellanox.com>
Cc: Shahaf Shuler <shahafs@mellanox.com>,
dev@dpdk.org, xuemingl@mellanox.com
Subject: Re: [dpdk-dev] [PATCH] net/mlx5: handle expected errno properly
Date: Wed, 29 Aug 2018 11:52:07 +0800 [thread overview]
Message-ID: <20180829035207.vlboi3og4o27rqvp@MTBC-JACKMIN.mtl.com> (raw)
In-Reply-To: <20180828204217.GA45628@yongseok-MBP.local>
On Tue, Aug 28, 2018 at 01:42:18PM -0700, Yongseok Koh wrote:
> On Fri, Aug 24, 2018 at 02:45:00PM +0800, Jack MIN wrote:
> > On Thu, Aug 23, 2018 at 02:08:09PM -0700, Yongseok Koh wrote:
> > > On Thu, Aug 23, 2018 at 02:38:51PM +0800, Xiaoyu Min wrote:
> > > > rte_errno is a per thread variable and is widely used as an
> > > > error indicator, which means a function could affect
> > > > other functions' results by setting rte_errno carelessly
> > > >
> > > > During rxq setup, an EINVAL rte_errno is expected since
> > > > the queues are not created yet
> > > > So rte_errno is cleared when it is EINVAL as expected
> > > >
> > > > Signed-off-by: Xiaoyu Min <jackmin@mellanox.com>
> > > > ---
> > > > drivers/net/mlx5/mlx5_rxq.c | 20 +++++++++++++++-----
> > > > 1 file changed, 15 insertions(+), 5 deletions(-)
> > > >
> > > > diff --git a/drivers/net/mlx5/mlx5_rxq.c b/drivers/net/mlx5/mlx5_rxq.c
> > > > index 1f7bfd4..e7056e8 100644
> > > > --- a/drivers/net/mlx5/mlx5_rxq.c
> > > > +++ b/drivers/net/mlx5/mlx5_rxq.c
> > > > @@ -443,6 +443,7 @@
> > > > struct mlx5_rxq_data *rxq = (*priv->rxqs)[idx];
> > > > struct mlx5_rxq_ctrl *rxq_ctrl =
> > > > container_of(rxq, struct mlx5_rxq_ctrl, rxq);
> > > > + int ret = 0;
> > > >
> > > > if (!rte_is_power_of_2(desc)) {
> > > > desc = 1 << log2above(desc);
> > > > @@ -459,13 +460,21 @@
> > > > rte_errno = EOVERFLOW;
> > > > return -rte_errno;
> > > > }
> > > > - if (!mlx5_rxq_releasable(dev, idx)) {
> > > > + ret = mlx5_rxq_releasable(dev, idx);
> > > > + if (!ret) {
> > > > DRV_LOG(ERR, "port %u unable to release queue index %u",
> > > > dev->data->port_id, idx);
> > > > rte_errno = EBUSY;
> > > > return -rte_errno;
> > > > + } else if (ret == -EINVAL) {
> > > > + /**
> > > > + * on the first time, rx queue doesn't exist,
> > > > + * so just ignore this error and reset rte_errno.
> > > > + */
> > > > + rte_errno = 0;
> > >
> > > Unless this function returns failure, the rte_errno will be ignored by caller
> > > and caller shouldn't assume rte_errno has 0. Caller will assume it is garbage
> > > data in case of success. So we can silently ignore this case. Does it cause any
> > > issue in application side?
> > >
> > Not application side but mlx5 PMD this time:
> > **mlx5_fdir_filter_delete**
> > which just _return -rte_errno;_
>
> Looks like an error. mlx5_fdir_filter_delete() can't be like that. We seem to
> have lost the code while refactoring it. Let take it offline.
>
Sure ~
> > For sure, _mlx5_fdir_filter_delete_ should be more defensive, should not assume
> > rte_errno is zero if no error happened.
> > However if the caller know that an error will happen and rte_errno will become
> > meaningless (garbage) for the succeeding functions, Catching the expected error
> > and resetting rte_errno will be better. What do you think?
>
> Still don't understand clearly. There would be many other similar cases where we
> don't clear rte_errno when returning success. I don't understand why this case
> should be taken as a special one??
>
No, the _mlx5_rxq_releasable()_ returns *error* (-EINVAL)(the rxq doesn't exist)
as my understanding
We only check if rxq is not releasable(==0) in previouse version but we didn't
check if function returns error or success
> Thanks
> Yongseok
next prev parent reply other threads:[~2018-08-29 3:52 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-08-23 6:38 Xiaoyu Min
2018-08-23 21:08 ` Yongseok Koh
2018-08-24 6:45 ` Jack MIN
2018-08-28 20:42 ` Yongseok Koh
2018-08-29 3:52 ` Jack MIN [this message]
2018-08-31 21:48 ` Yongseok Koh
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20180829035207.vlboi3og4o27rqvp@MTBC-JACKMIN.mtl.com \
--to=jackmin@mellanox.com \
--cc=dev@dpdk.org \
--cc=shahafs@mellanox.com \
--cc=xuemingl@mellanox.com \
--cc=yskoh@mellanox.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).