On Sun, Dec 30, 2001 at 05:01:41PM +0200, Sheldon Hearn wrote:
> On Sun, 30 Dec 2001 15:47:45 +0100, Michal Mertl wrote:
>
> > I did use the "goto oldway;" and the problem went away. I tried to look at
> > /sys/kern/uipc_syscalls.c sendfile implementation but it is too complex
> > for me :-(.
>
> I've added your feedback to the audit trail of PR bin/31692, thanks.
>
> > I tried ktrace on ftpd but only saw the call to sendfile(2). If you
> > give me some guidance I can try to look into problem deeper. I don't
> > have any experience in kernel debugging but would like to learn it.
>
> No idea. I was just trying to narrow this down so we know who to nag.
>
> This smells like one of those things someone like dillon or bmilekic
> could fix. Gents, any takers?
I don't know if this is correct, but I find the usage of cnt for
the loop to be very suspicious. Could you try this patch (against
-CURRENT, but should apply to -STABLE)?
Consider that sendfile may be executed many times and cnt will only
be the value of how much was written that last iteration. Therefore,
you need to check whether the *entire* file has been written. And, I
would wonder with a network-based FS, it would potentially take
multiple sendfile(2) calls - therefore the check against cnt is
invalid.
And, of course, ftpd.c isn't built against libc_r. Otherwise,
I'd say to upgrade your kernel. Alfred got the libc_r sendfile
working again a few weeks ago. -- justin
--- libexec/ftpd/ftpd.c.old Mon Nov 19 13:52:03 2001
+++ libexec/ftpd/ftpd.c Sun Dec 30 11:33:26 2001
@@ -1811,7 +1811,7 @@
len = filesize;
err = cnt = offset = 0;
- while (err != -1 && cnt < filesize) {
+ while (err != -1 && len > 0) {
err = sendfile(filefd, netfd, offset, len,
(struct sf_hdtr *) NULL, &cnt, 0);
byte_count += cnt;
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-current" in the body of the message
