DJ Lucas wrote:
Yes at least the echo is required.
Get to that in a moment. The loop is as well, but it's still undecided
according to the comments in the ticket. I'll be able to test at least
the loop doesn't do anything silly tonight, but I don't believe I have a
reliable way to reproduce the bug as my system is now.
I treat the ticket 1720 as presumably-closed if the bootscript changes
below are implemented, and the udev_update branch can be merged then.
This mail to lfs-dev is only to state this, and ask for discussion if
such discussion is needed (IMHO, it is not needed).
The status quo is:
1) Uevents that are triggered by echoing to sysfs "uevent" files are
processed by udev asynchronously. The end of the loop that does the
"echo" only means that the kernel has put all uevents into the netlink
socket buffer. There is absolutely no guarantee that udevd received them
all at that moment.
2) Therefore, upstream has the recommendation to wait for udevd to
process all these events, and all uevents raised as a consequence of
"modprobe" called from udev rules in response to earlier uevents.
Upstream provides a way to know if the udev queue (that is not the same
as in-kernel netlink socket buffer!) is empty, and proposes the
following shell script snippet:
udevd --daemon
mkdir -p /dev/.udev/queue
walk_sysfs
# until we know how to do better, just wait for _all_ events to finish
loop=300
while test -d /dev/.udev/queue; do
sleep 0.1
test "$loop" -gt 0 || break
loop=$(($loop - 1))
done
test "$loop" -gt 0
evaluate_retval
3) The shell script above does not work. Udev queue disappears for a
moment (because udevd has no way to know if there will be any further
uevents), and then reappears, but the loop thinks that it is done! So I
proposed an alternative loop that not only waits for the queue to
disappear, but retries several times in order to make sure that it
doesn't reappear again:
udevd --daemon
mkdir -p /dev/.udev/queue
walk_sysfs
# until we know how to do better, just wait for _all_ events to finish
loop=300
confirm=0
while true ; do
sleep 0.1
test -d /dev/.udev/queue && confirm=0 || \
confirm=$(( $confirm + 1 ))
loop=$(( $loop - 1 ))
test $loop -gt 0 || break
test $confirm -lt 60 || break
done
test "$loop" -gt 0
evaluate_retval
Archaic says this produces much better results on his system.
4) It is imperative that we don't trust upstream (because they ignore
problem (3)), and do our own checking. In order to make sure that no
initial uevents escape the loop, compile the following simple event
recorder that adds the uevent contents to the /dev/bug file if the file
exists:
/* bug.c: Simple event recorder, gcc -o /lib/udev/bug bug.c */
#define _GNU_SOURCE
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <unistd.h>
#include <stdlib.h>
#include <argz.h>
int main(int argc, char * argv[])
{
char * envz;
size_t len;
int bug;
bug = open("/dev/bug", O_WRONLY | O_APPEND);
if (bug == -1)
return 0;
setenv("_SEPARATOR", "------------------------------------", 1);
argz_create(environ, &envz, &len);
argz_stringify(envz, len, '\n');
envz[len-1]='\n';
write(bug, envz, len);
close(bug);
free(envz);
return 0;
}
Add this rule to the separate file, 90-debug.rules:
ACTION=="add", RUN+="bug"
Add this shell script snippet to the udev bootscripts after the waiting
loop. This snippt checks if there are any uevents within 6 seconds from
the end of the loop. Assuming that the user doesn't plug devices in and
out during the boot process, such uevents indicate a bug in the waiting
loop.
.....
test $confirm -lt 60 || break
done
>/dev/bug
test "$loop" -gt 0
evaluate_retval
sleep 5
if test -s /dev/bug; then
mv /dev/bug /dev/bugreport
boot_mesg "Please paste the /dev/bugreport file to" ${WARNING}
boot_mesg "http://wiki.linuxfromscratch.org/lfs/ticket/1720";
boot_mesg "Otherwise, the next version of LFS may be
unbootable on your system!"
echo_failure
sleep 10
else
rm -f /dev/bug
fi
--
Alexander E. Patrakov
--
http://linuxfromscratch.org/mailman/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page
