From: Tyler Retzlaff <roretzla@linux.microsoft.com>
To: Long Li <longli@microsoft.com>
Cc: Ferruh Yigit <ferruh.yigit@amd.com>,
stephen <stephen@networkplumber.org>,
Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>,
"dev@dpdk.org" <dev@dpdk.org>
Subject: Re: [Patch v2] net/mana: use rte_pktmbuf_alloc_bulk for allocating RX WQEs
Date: Thu, 1 Feb 2024 08:33:05 -0800 [thread overview]
Message-ID: <20240201163305.GB13514@linuxonhyperv3.guj3yctzbm1etfxqx2vob5hsef.xx.internal.cloudapp.net> (raw)
In-Reply-To: <MN0PR21MB32640C60E08795B35881DE9BCE432@MN0PR21MB3264.namprd21.prod.outlook.com>
On Thu, Feb 01, 2024 at 03:55:55AM +0000, Long Li wrote:
> > >> 'mbufs' is temporarily storage for allocated mbuf pointers, why not
> > >> allocate if from stack instead, can be faster and easier to manage:
> > >> "struct rte_mbuf *mbufs[count]"
> > >
> > > That would introduce a variable length array.
> > > VLA's should be removed, they are not supported on Windows and many
> > > security tools flag them. The problem is that it makes the code
> > > brittle if count gets huge.
> > >
> > > But certainly regular calloc() or alloca() would work here.
> > >
> >
> > Most of the existing bulk alloc already uses VLA but I can see the problem it is not
> > being supported by Windows.
> >
> > As this mbuf pointer array is short lived within the function, and being in the fast
> > path, I think continuous alloc and free can be prevented,
> >
> > one option can be to define a fixed size, big enough, array which requires
> > additional loop for the cases 'count' size is bigger than array size,
> >
> > or an array can be allocated by driver init in device specific data ,as we know it
> > will be required continuously in the datapath, and it can be freed during device
> > close()/uninit().
> >
> > I think an fixed size array from stack is easier and can be preferred.
>
> I sent a v3 of the patch, still using alloc().
>
> I found two problems with using a fixed array:
> 1. the array size needs to be determined in advance. I don't know what a good number should be. If too big, some of them may be wasted. (and maybe make a bigger mess of CPU cache) If too small, it ends up doing multiple allocations, which is the problem this patch trying to solve.
> 2. if makes the code slightly more complex ,but I think 1 is the main problem.
>
> I think another approach is to use VLA by default, but for Windows use alloc().
a few thoughts on VLAs you may consider. not to be regarded as a strong
objection.
indications are that standard C will gradually phase out VLAs because
they're generally accepted as having been a bad idea. that said
compilers that implement them will probably keep them forever.
VLAs generate a lot of code relative to just using a more permanent
allocation. may not show up in your performance tests but you also may
not want it on your hotpath either.
mana doesn't currently support windows, are there plans to support
windows? if never then i suppose VLAs can be used since all the
toolchains you care about have them. though it does raise the bar, cause
more work, later refactor, carry regression risk should you change your
mind and choose to port to windows.
accepting the use of VLAs anywhere in dpdk prohibits general
checkpatches and/or compiling with compiler options that detect and flag
their inclusion as a part of the CI without having to add exclusion
logic for drivers that are allowed to use them.
>
> Long
next prev parent reply other threads:[~2024-02-01 16:33 UTC|newest]
Thread overview: 25+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-01-25 2:42 [PATCH] " longli
2024-01-26 0:29 ` Stephen Hemminger
2024-01-26 1:13 ` Long Li
2024-01-30 1:13 ` [Patch v2] " longli
2024-01-30 10:19 ` Ferruh Yigit
2024-01-30 16:43 ` Stephen Hemminger
2024-01-30 18:05 ` Tyler Retzlaff
2024-01-30 22:42 ` Ferruh Yigit
2024-02-01 3:55 ` Long Li
2024-02-01 10:52 ` Ferruh Yigit
2024-02-02 1:21 ` Long Li
2024-02-01 16:33 ` Tyler Retzlaff [this message]
2024-02-02 1:22 ` Long Li
2024-01-30 21:30 ` Long Li
2024-01-30 22:34 ` Ferruh Yigit
2024-01-30 22:36 ` Long Li
2024-02-01 3:45 ` [Patch v3] " longli
2024-02-01 16:16 ` Tyler Retzlaff
2024-02-01 19:41 ` Long Li
2024-02-02 1:19 ` [Patch v4] net/mana: use rte_pktmbuf_alloc_bulk for allocating RX mbufs longli
2024-02-02 16:24 ` Stephen Hemminger
2024-02-06 18:06 ` Ferruh Yigit
2024-02-07 4:50 ` Long Li
2024-02-09 0:02 ` [Patch v5] net/mana: use rte_pktmbuf_alloc_bulk for allocating RX WQEs longli
2024-02-09 17:46 ` Ferruh Yigit
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=20240201163305.GB13514@linuxonhyperv3.guj3yctzbm1etfxqx2vob5hsef.xx.internal.cloudapp.net \
--to=roretzla@linux.microsoft.com \
--cc=andrew.rybchenko@oktetlabs.ru \
--cc=dev@dpdk.org \
--cc=ferruh.yigit@amd.com \
--cc=longli@microsoft.com \
--cc=stephen@networkplumber.org \
/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).