[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <Y+ANVTMT7jgV0i0l@DEN-LT-70577>
Date: Sun, 5 Feb 2023 20:11:02 +0000
From: <Daniel.Machon@...rochip.com>
To: <simon.horman@...igine.com>
CC: <netdev@...r.kernel.org>, <davem@...emloft.net>,
<edumazet@...gle.com>, <kuba@...nel.org>, <pabeni@...hat.com>,
<Lars.Povlsen@...rochip.com>, <Steen.Hegelund@...rochip.com>,
<UNGLinuxDriver@...rochip.com>, <joe@...ches.com>,
<richardcochran@...il.com>, <casper.casan@...il.com>,
<Horatiu.Vultur@...rochip.com>, <shangxiaojing@...wei.com>,
<rmk+kernel@...linux.org.uk>, <nhuck@...gle.com>,
<error27@...il.com>, <linux-kernel@...r.kernel.org>,
<linux-arm-kernel@...ts.infradead.org>
Subject: Re: [PATCH net-next 03/10] net: microchip: sparx5: add support for
Service Dual Leacky Buckets
Hi Simon,
Thanks for reviewing my patches, appreciate it!
> > +static u32 sparx5_sdlb_group_get_last(struct sparx5 *sparx5, u32 group)
> > +{
> > + u32 itr, next;
> > +
> > + itr = sparx5_sdlb_group_get_first(sparx5, group);
> > +
> > + for (;;) {
>
> Unbounded loops like this give me some apprehension.
> Will they always terminate?
Yes, it will always terminate - unless the add() del() functions are
buggy to begin with.
The end of the leak chain is marked by an index pointing to itself, and
this is the exit condition I am looking for in the unbounded loop.
>
> > + next = sparx5_sdlb_group_get_next(sparx5, group, itr);
> > + if (itr == next)
> > + return itr;
> > +
> > + itr = next;
> > + }
> > +}
>
> ...
>
> > +static int sparx5_sdlb_group_get_adjacent(struct sparx5 *sparx5, u32 group,
> > + u32 idx, u32 *prev, u32 *next,
> > + u32 *first)
> > +{
> > + u32 itr;
> > +
> > + *first = sparx5_sdlb_group_get_first(sparx5, group);
> > + *prev = *first;
> > + *next = *first;
> > + itr = *first;
> > +
> > + for (;;) {
> > + *next = sparx5_sdlb_group_get_next(sparx5, group, itr);
> > +
> > + if (itr == idx)
> > + return 0; /* Found it */
> > +
> > + if (itr == *next)
> > + return -EINVAL; /* Was not found */
> > +
> > + *prev = itr;
> > + itr = *next;
> > + }
> > +}
> > +
> > +static int sparx5_sdlb_group_get_count(struct sparx5 *sparx5, u32 group)
> > +{
> > + u32 itr, next;
> > + int count = 0;
> > +
> > + itr = sparx5_sdlb_group_get_first(sparx5, group);
> > +
> > + for (;;) {
> > + next = sparx5_sdlb_group_get_next(sparx5, group, itr);
> > + if (itr == next)
> > + return count;
> > +
> > + itr = next;
> > + count++;
> > + }
> > +}
>
> ...
>
> > +int sparx5_sdlb_group_get_by_index(struct sparx5 *sparx5, u32 idx, u32 *group)
> > +{
> > + u32 itr, next;
> > + int i;
> > +
> > + for (i = 0; i < SPX5_SDLB_GROUP_CNT; i++) {
> > + if (sparx5_sdlb_group_is_empty(sparx5, i))
> > + continue;
> > +
> > + itr = sparx5_sdlb_group_get_first(sparx5, i);
> > +
> > + for (;;) {
> > + next = sparx5_sdlb_group_get_next(sparx5, i, itr);
> > +
> > + if (itr == idx) {
> > + *group = i;
> > + return 0; /* Found it */
> > + }
> > + if (itr == next)
> > + break; /* Was not found */
> > +
> > + itr = next;
> > + }
> > + }
> > +
> > + return -EINVAL;
> > +}
>
> ...
Powered by blists - more mailing lists