From: Jeff Mahoney <jeffm@suse.com>
To: Christoph Hellwig <hch@infradead.org>
Cc: Linux SCSI Mailing List <linux-scsi@vger.kernel.org>
Subject: Re: [PATCH] scsi: iterate over devices individually for /proc/scsi/scsi
Date: Mon, 29 Oct 2007 17:50:32 -0400 [thread overview]
Message-ID: <472655A8.1090700@suse.com> (raw)
In-Reply-To: <20070413072506.GA20326@infradead.org>
Christoph Hellwig wrote:
> On Thu, Apr 12, 2007 at 03:34:32PM -0400, Jeff Mahoney wrote:
>> On systems with very large numbers (> 1600 or so) of SCSI devices,
>> cat /proc/scsi/scsi ends up failing with -ENOMEM. This is due to
>> the show routine simply iterating over all of the devices with
>> bus_for_each_dev(), and trying to dump all of them into the buffer
>> at the same time. On my test system (using scsi_debug with 4064 devices),
>> the output ends up being ~ 632k, far more than kmalloc will typically allow.
>>
>> This patch uses seq_file directly instead of single_file, and breaks up
>> the operations into the 4 seq_file callbacks. The result is that
>> each show() operation only dumps ~ 180 bytes into the buffer at a time
>> so we don't run out of memory.
>
> The simpler answer is don't use /proc/scsi/scsi, but lsscsi instead.
> It even has an option to provide /proc/scsi/scsi-style output.
>
> This already came up the last time someone from SuSE sent a patch
> to fix an "issue" like that. That time at least the patch was
> massageable into something sane, which this one isn't. So please
> take the advice this time and stop using /proc/scsi/scsi.
This should be something a bit more sane. After writing this up, I'm in
complete agreement with you on how much of an unsalvagable mess the
previous patch was.
On systems with very large numbers (> 1600 or so) of SCSI devices,
cat /proc/scsi/scsi ends up failing with -ENOMEM. This is due to
the show routine simply iterating over all of the devices with
bus_for_each_dev(), and trying to dump all of them into the buffer
at the same time. On my test system (using scsi_debug with 4064 devices),
the output ends up being ~ 632k, far more than kmalloc will typically allow.
This patch defines its own seq_file opreations to iterate over the scsi
devices.The result is that each show() operation only dumps ~ 180 bytes
into the buffer at a time so we don't run out of memory.
If the "Attached devices" header isn't required, we can dump the
sfile->private bit completely.
Signed-off-by: Jeff Mahoney <jeffm@suse.com>
---
drivers/scsi/scsi_proc.c | 58 ++++++++++++++++++++++++++++++++++++++++++-----
1 file changed, 52 insertions(+), 6 deletions(-)
--- a/drivers/scsi/scsi_proc.c 2007-10-29 13:56:28.000000000 -0400
+++ b/drivers/scsi/scsi_proc.c 2007-10-29 15:50:57.000000000 -0400
@@ -294,20 +294,66 @@ static ssize_t proc_scsi_write(struct fi
return err;
}
-static int proc_scsi_show(struct seq_file *s, void *p)
+static int always_match(struct device *dev, void *data)
{
- seq_printf(s, "Attached devices:\n");
- bus_for_each_dev(&scsi_bus_type, NULL, s, proc_print_scsidevice);
- return 0;
+ return 1;
+}
+
+static inline struct device *next_scsi_device(struct device *start)
+{
+ struct device *next = bus_find_device(&scsi_bus_type, start, NULL,
+ always_match);
+ put_device(start);
+ return next;
+}
+
+static void *scsi_seq_start(struct seq_file *sfile, loff_t *pos)
+{
+ struct device *dev = NULL;
+ loff_t n = *pos;
+
+ while ((dev = next_scsi_device(dev))) {
+ if (!n--)
+ break;
+ sfile->private++;
+ }
+ return dev;
+}
+
+static void *scsi_seq_next(struct seq_file *sfile, void *v, loff_t *pos)
+{
+ (*pos)++;
+ sfile->private++;
+ return next_scsi_device(v);
+}
+
+static void scsi_seq_stop(struct seq_file *sfile, void *v)
+{
+ put_device(v);
+}
+
+static int scsi_seq_show(struct seq_file *sfile, void *dev)
+{
+ if (!sfile->private)
+ seq_puts(sfile, "Attached devices:\n");
+
+ return proc_print_scsidevice(dev, sfile);
}
+static struct seq_operations scsi_seq_ops = {
+ .start = scsi_seq_start,
+ .next = scsi_seq_next,
+ .stop = scsi_seq_stop,
+ .show = scsi_seq_show
+};
+
static int proc_scsi_open(struct inode *inode, struct file *file)
{
/*
* We don't really needs this for the write case but it doesn't
* harm either.
*/
- return single_open(file, proc_scsi_show, NULL);
+ return seq_open(file, &scsi_seq_ops);
}
static struct file_operations proc_scsi_operations = {
@@ -315,7 +361,7 @@ static struct file_operations proc_scsi_
.read = seq_read,
.write = proc_scsi_write,
.llseek = seq_lseek,
- .release = single_release,
+ .release = seq_release,
};
int __init scsi_init_procfs(void)
--
Jeff Mahoney
SUSE Labs
next prev parent reply other threads:[~2007-10-29 21:47 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2007-04-12 19:34 [PATCH] scsi: iterate over devices individually for /proc/scsi/scsi Jeff Mahoney
2007-04-12 22:01 ` Jeff Mahoney
2007-04-13 7:25 ` Christoph Hellwig
2007-10-29 21:50 ` Jeff Mahoney [this message]
2007-10-30 10:17 ` Christoph Hellwig
-- strict thread matches above, loose matches on Subject: below --
2011-04-27 20:22 Jeff Mahoney
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=472655A8.1090700@suse.com \
--to=jeffm@suse.com \
--cc=hch@infradead.org \
--cc=linux-scsi@vger.kernel.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).