[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <1300758651.21880.207.camel@haakon2.linux-iscsi.org>
Date: Mon, 21 Mar 2011 18:50:51 -0700
From: "Nicholas A. Bellinger" <nab@...ux-iscsi.org>
To: Mike Christie <michaelc@...wisc.edu>
Cc: linux-scsi <linux-scsi@...r.kernel.org>,
linux-kernel <linux-kernel@...r.kernel.org>,
James Bottomley <James.Bottomley@...senPartnership.com>,
Christoph Hellwig <hch@....de>, Hannes Reinecke <hare@...e.de>,
FUJITA Tomonori <fujita.tomonori@....ntt.co.jp>,
Boaz Harrosh <bharrosh@...asas.com>,
Stephen Rothwell <sfr@...b.auug.org.au>,
Andrew Morton <akpm@...ux-foundation.org>,
Douglas Gilbert <dgilbert@...erlog.com>,
Jesper Juhl <jj@...osbits.net>
Subject: Re: [RFC-v4 07/12] iscsi-target: Add CHAP Authentication support
using libcrypto
On Mon, 2011-03-21 at 20:05 -0500, Mike Christie wrote:
> On 03/20/2011 04:31 AM, Nicholas A. Bellinger wrote:
> > +static struct iscsi_chap *chap_server_open(
> > + struct iscsi_conn *conn,
> > + struct iscsi_node_auth *auth,
> > + const char *A_str,
> > + char *AIC_str,
> > + unsigned int *AIC_len)
>
> Lot of mixed cases like this and below in the patch.
>
> > +static int chap_server_compute_md5(
> > + struct iscsi_conn *conn,
> > + struct iscsi_node_auth *auth,
> > + char *NR_in_ptr,
> > + char *NR_out_ptr,
> > + unsigned int *NR_out_len)
>
>
<nod>, fixing this up in lio-4.1 now..
>
> > + return 2;
> > +}
> > diff --git a/drivers/target/iscsi/iscsi_target_auth.h b/drivers/target/iscsi/iscsi_target_auth.h
> > new file mode 100644
> > index 0000000..17b042d
> > --- /dev/null
> > +++ b/drivers/target/iscsi/iscsi_target_auth.h
>
>
>
> > +
> > +struct iscsi_chap {
> > + unsigned char digest_type;
> > + unsigned char id;
> > + unsigned char challenge[CHAP_CHALLENGE_LENGTH];
> > + unsigned int challenge_len;
> > + unsigned int authenticate_target;
> > + unsigned int chap_state;
> > +} ____cacheline_aligned;
> > +
>
> Why are almost all structs in the patches ____cacheline_aligned? Is it
> something we are just doing now, or does this affect performance somehow
> even though the struct is not used in a perf path?
For this particular case I wanted to make sure it was at least 32-bit
word aligned for ARM to avoid the unaligned penalites for the single
byte unsigned chars.
There is also a handful of single byte unsigned chars, u8, and u16
structure membmer usage in iscsi_target_core.h that is using
____cacheline_aligned using in the non performance critical path.
I would be happy to do a audit on these and convert to using the packed
attributed..
Just to double check, what is mainline convention for handling
structures like this for 32-bit word aligned ISAs..?
Thanks for your review Mike!
--nab
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists