Re: CRC32 - computed or table-driven?

From: Matt_Domsch@Dell.com
Date: Mon May 06 2002 - 15:55:23 EST


"Stanislav Meduna" <stano@meduna.org> wrote:
> I need to compute a crc32 in my small not-ready-for-the-prime-time
> patch against 2.4.19-pre. I do not (yet) follow the 2.5 development.
>
> There is <linux/crc32.h> that does what I want, but not quite
> effectively and with comments that it is unsuitable for bulk data
> and that it will migrate to net/core/crc.c.

Right. In 2.5, there's a bulk data version, but that code hasn't been
backported to 2.4.x, and that was so far intentional. (the comment about
net/core/crc.c is wrong, it should eventually go in lib/crc32.c).

> Any plans to do it "the right way" in 2.4?

Yes. The stage 1 crc32 cleanups I did just removed ~60 duplicate copies of
the crc32 code in the ethernet drivers, but we (Jeff Garzik and Marcello)
didn't want more widespread changes. That all happened in 2.4.19-pre{early}
IIRC. Once they were convinced that the crc library code in 2.5.x wouldn't
break the drivers, it was believed that the code could easily be backported
- but no one has volunteered to do so yet. Stage 2 is to add lib/crc32.c as
a module that adds the bulk crc32() function, but doesn't touch
ether_crc_le() or ether_crc(), so that can't break net drivers.

> I can post a patch, but I cannot test whether it breaks jffs2
> and as this adds 1 kB to the kernel that might not be needed
> I would like to ask first. Who is hacking these parts?

If you wanted to backport from 2.5 the lib/crc32.c code that generates the
table and exports a crc32 function, be my guest!

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



This archive was generated by hypermail 2b29 : Tue May 07 2002 - 22:00:28 EST