From: Christoph Hellwig <hch@infradead.org>
To: Stefan Richter <stefanr@s5r6.in-berlin.de>
Cc: Christoph Hellwig <hch@infradead.org>,
linux-kernel@vger.kernel.org, Kristian H??gsberg <krh@redhat.com>,
Linus Torvalds <torvalds@linux-foundation.org>,
Andrew Morton <akpm@linux-foundation.org>,
linux1394-devel <linux1394-devel@lists.sourceforge.net>,
jejb@steeleye.com
Subject: Re: [PATCH 5/6] firewire: SBP-2 highlevel driver
Date: Fri, 4 May 2007 10:53:10 +0100 [thread overview]
Message-ID: <20070504095310.GB31811@infradead.org> (raw)
In-Reply-To: <4639085A.8010504@s5r6.in-berlin.de>
On Wed, May 02, 2007 at 11:53:30PM +0200, Stefan Richter wrote:
> > Please set the max_sectors value in your host template so that the
> > block layer doesn't build sg entries too big for you.
>
> Hmm, what about this:
>
> James Bottomley wrote on 2007-01-15:
> | Actually, there's one unfortunate case where Linux won't respect this:
> | an IOMMU that can do virtual merging. This parameter is a block queue
> | parameter, so block will happily make sure the request segments obey it.
> | However, when you get to dma_map_rq() it doesn't see the segment limits,
> | so, if the iommu merges, you can end up with SG elements the other side
> | that violate this. I've been meaning to do something about this for
> | ages (IDE is the other subsystem that has an absolute requirement for a
> | fixed maximum segment size) but never found an excuse to fix it.
>
> http://marc.info/?l=linux-scsi&m=116889641203397
Hmm, okay. Probably wants a comment in there explaining it, and we
should poke jejb to fix it for real :)
> >> +static int add_scsi_devices(struct fw_unit *unit)
> >> +{
> >> + struct sbp2_device *sd = unit->device.driver_data;
> >> + int retval, lun;
> >> +
> >> + if (sd->scsi_host != NULL)
> >> + return 0;
> >> +
> >> + sd->scsi_host = scsi_host_alloc(&scsi_driver_template,
> >> + sizeof(unsigned long));
> >> + if (sd->scsi_host == NULL) {
> >> + fw_error("failed to register scsi host\n");
> >> + return -1;
> >> + }
> >> +
> >> + sd->scsi_host->hostdata[0] = (unsigned long)unit;
> >
> > Please take a look ar ther other scsi drivers how this is supposed
> > to be used.
>
> Do you mean the one Scsi_Host per LU? If it is that, then it was just
> taken over from drivers/ieee1394/sbp2.c. Sbp2 is doing this still today
> mostly for historical reasons; I just didn't find the time yet to try to
> get to a leaner scheme.
No, sorry. I should have written a better explanation. scsi_host_alloc
is designed to allocate space for your private data aswell. So you
should call it early on an allocate the sbp2_device as part of the Scsi_Host
instead of just stuffing in a pointer.
> > Do we really need another scanning algorithm?
>
> Yes.
>
> > Can't you use scsi_scan_target instead and let the core scsi code
> > handle the scanning?
>
> No. The discovery of LUs of SBP-2 targets happens on the IEEE 1212
> level of things. The initiator has to parse the configuration ROM of
> the target FireWire node; the ROM has entries for each LU. (After that,
> SBP-2 login protocol commences for each LU, and only after that can SCSI
> requests be issued. There is nothing SCSIish going on before that.)
>
> What's missing as a /* FIXME */ here is actually implemented in the
> mainline sbp2.c and needs to be brought over here; converted to the new
> FireWire core APIs.
Okay, so sbp2 decided to be non-standard here, what a pity. It's probably
better to use scsi_scan_target with a specific lun, though as scsi_add_device
is a rather awkward API.
> > This function seems rather oddly named. And the checking and
> > setting of scsi_host looks like you have some lifetime rule
> > problems.
> >
>
> The NULL probably has to do with the ability to call remove_scsi_devices
> in different paths. (These paths are not concurrent.)
Needs documentation at least. And at least my preference would be
to have a deleted flag instead of the null setting because the latter
can easily paper over bugs.
next prev parent reply other threads:[~2007-05-04 9:53 UTC|newest]
Thread overview: 59+ messages / expand[flat|nested] mbox.gz Atom feed top
2007-05-01 20:27 [git pull] New firewire stack Kristian Høgsberg
2007-05-01 21:34 ` Stefan Richter
2007-05-02 9:00 ` Christoph Hellwig
2007-05-02 12:13 ` Stefan Richter
2007-05-02 12:15 ` [PATCH 1/6] firewire: handling of cards, buses, nodes Stefan Richter
2007-05-02 12:16 ` [PATCH 2/6] firewire: isochronous and asynchronous I/O Stefan Richter
2007-05-02 12:17 ` [PATCH 3/6] firewire: char device interface Stefan Richter
2007-05-02 12:18 ` [PATCH 4/6] firewire: OHCI-1394 lowlevel driver Stefan Richter
2007-05-02 12:18 ` [PATCH 5/6] firewire: SBP-2 highlevel driver Stefan Richter
2007-05-02 12:19 ` [PATCH 6/6] firewire: add it all to kbuild Stefan Richter
2007-05-02 18:05 ` Stefan Richter
2007-05-02 19:44 ` Christoph Hellwig
2007-05-02 23:01 ` Stefan Richter
2007-05-03 4:15 ` Sam Ravnborg
2007-05-03 8:10 ` Christoph Hellwig
2007-05-08 0:14 ` Kristian Høgsberg
2007-05-02 19:44 ` [PATCH 5/6] firewire: SBP-2 highlevel driver Christoph Hellwig
2007-05-02 21:53 ` Stefan Richter
2007-05-02 22:10 ` Stefan Richter
2007-05-04 9:53 ` Christoph Hellwig [this message]
2007-05-04 11:20 ` Stefan Richter
2007-05-09 21:05 ` Kristian Høgsberg
2007-05-09 21:48 ` Stefan Richter
2007-05-09 21:57 ` Stefan Richter
2007-05-09 22:13 ` Kristian Høgsberg
2007-05-09 22:56 ` Stefan Richter
2007-05-04 11:11 ` [PATCH 4/6] firewire: OHCI-1394 lowlevel driver Christoph Hellwig
2007-05-09 23:40 ` Kristian Høgsberg
2007-05-02 15:35 ` [PATCH 3/6] firewire: char device interface John Stoffel
2007-05-02 16:06 ` Stefan Richter
2007-05-02 21:11 ` Kristian Høgsberg
2007-05-04 9:48 ` Christoph Hellwig
2007-05-08 0:19 ` Kristian Høgsberg
2007-05-02 19:30 ` Christoph Hellwig
2007-05-08 0:08 ` Kristian Høgsberg
2007-05-02 19:29 ` [PATCH 2/6] firewire: isochronous and asynchronous I/O Christoph Hellwig
2007-05-03 0:08 ` Kristian Høgsberg
2007-05-03 8:54 ` Stefan Richter
2007-05-02 15:55 ` [PATCH 1/6] firewire: handling of cards, buses, nodes Pekka Enberg
2007-05-02 19:16 ` Stefan Richter
2007-05-02 20:35 ` Pekka Enberg
2007-05-07 22:02 ` Kristian Høgsberg
2007-05-02 21:16 ` Kristian Høgsberg
2007-05-02 19:22 ` Christoph Hellwig
2007-05-07 23:42 ` Kristian Høgsberg
2007-05-02 20:00 ` [git pull] New firewire stack Kristian Høgsberg
2007-05-02 12:21 ` Olaf Hering
2007-05-02 12:48 ` Stefan Richter
2007-05-02 13:56 ` Gene Heskett
2007-05-02 18:51 ` Stefan Richter
2007-05-02 15:27 ` Adrian Bunk
2007-05-02 20:03 ` Kristian Høgsberg
2007-05-02 19:53 ` Kristian Høgsberg
2007-05-02 20:03 ` Olaf Hering
2007-05-10 17:26 ` [git pull] New firewire stack (updated) Stefan Richter
2007-05-10 17:38 ` Christoph Hellwig
2007-05-10 17:51 ` Adrian Bunk
2007-05-10 17:56 ` Stefan Richter
2007-05-10 18:05 ` Stefan Richter
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=20070504095310.GB31811@infradead.org \
--to=hch@infradead.org \
--cc=akpm@linux-foundation.org \
--cc=jejb@steeleye.com \
--cc=krh@redhat.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux1394-devel@lists.sourceforge.net \
--cc=stefanr@s5r6.in-berlin.de \
--cc=torvalds@linux-foundation.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
all inboxes | Powered by JetHome®