Skip to contentSkip to conversation

Udev incompatibility with qemu and bochs

DISCUSSION Alexander E. Patrakov 3 participants 4 messages Oct 19, 2005 – Oct 20, 2005

lore
Share & tools
lore
#0 lore raw

Hello,

udevstart contains the following code:

static void asmlinkage sig_handler(int signum)

{
        switch (signum) {
                case SIGALRM:
                        exit(1);
                case SIGINT:
                case SIGTERM:
                        exit(20 + signum);
        }
}

int main(int argc, char *argv[], char *envp[])
{
/* snip */

        sigaction(SIGALRM, &act, NULL);
        sigaction(SIGINT, &act, NULL);
        sigaction(SIGTERM, &act, NULL);

        /* trigger timeout to prevent hanging processes */
        alarm(UDEV_ALARM_TIMEOUT);

/* snip */
}

UDEV_ALARM_TIMEOUT is defined to 120 seconds in udev.h. The problem is that it does take 2 minutes to populate /dev inside slow emulators like qemu or bochs, especially when the emulated and host CPU types differ and udev rules include RUN+=/sbin/udev_run_hotplugd. The reason is that inside those emulators exec() is very expensive (involves jit-compilation from one sort of machine code to another). Please provide an option to disable this alarm. The preferred form would be an environment variable, so that I can type "linux UDEV_NO_ALARM=1" at the boot prompt inside the emulator and see the system booting normally.

Thanks,

Signature · 11 lines
Alexander E. Patrakov


-------------------------------------------------------
This SF.Net email is sponsored by:
Power Architecture Resource Center: Free content, downloads, discussions,
and more. http://solutions.newsforge.com/ibmarch.tmpl
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel
#1 lore raw
Quoted 9 lines · Alexander E. Patrakov

UDEV_ALARM_TIMEOUT is defined to 120 seconds in udev.h. The problem is that it does take 2 minutes to populate /dev inside slow emulators like qemu or bochs, especially when the emulated and host CPU types differ and udev rules include RUN+=/sbin/udev_run_hotplugd. The reason is that inside those emulators exec() is very expensive (involves jit-compilation from one sort of machine code to another). Please provide an option to disable this alarm. The preferred form would be an environment variable, so that I can type "linux UDEV_NO_ALARM=1" at the boot prompt inside the emulator and see the system booting normally.

Feel free to provide such a patch, if you need it.

thanks,

greg k-h

-------------------------------------------------------
This SF.Net email is sponsored by:
Power Architecture Resource Center: Free content, downloads, discussions, and more. http://solutions.newsforge.com/ibmarch.tmpl
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#2 lore raw
In reply to Alexander E. Patrakov
Quoted 11 lines · Greg KH
Quoted 9 lines · Alexander E. Patrakov

UDEV_ALARM_TIMEOUT is defined to 120 seconds in udev.h. The problem is that it does take 2 minutes to populate /dev inside slow emulators like qemu or bochs, especially when the emulated and host CPU types differ and udev rules include RUN+=/sbin/udev_run_hotplugd. The reason is that inside those emulators exec() is very expensive (involves jit-compilation from one sort of machine code to another). Please provide an option to disable this alarm. The preferred form would be an environment variable, so that I can type "linux UDEV_NO_ALARM=1" at the boot prompt inside the emulator and see the system booting normally.

Quoted 1 line

Feel free to provide such a patch, if you need it.

This timeout is a leftover from the days of tdb, which didn't work well and caused udev processes to spin around in a corrupted database.

I generally don't like to introduce more facy options to udev and better set the alarm timeout to a reasonable default or remove it at all.

The problem occurs only on bootup running udevstart, right? (Depending on your setup 'UDEV_RUN=0 /sbin/udevstart' may be a good option to speed up the initial population of /dev. Don't know it you really need to have RUN at that time, depends on how udev is integrated.)

Kay

-------------------------------------------------------
This SF.Net email is sponsored by:
Power Architecture Resource Center: Free content, downloads, discussions, and more. http://solutions.newsforge.com/ibmarch.tmpl
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#3 lore raw
In reply to Alexander E. Patrakov
Quoted 7 lines · Kay Sievers

This timeout is a leftover from the days of tdb, which didn't work well and caused udev processes to spin around in a corrupted database.

I generally don't like to introduce more facy options to udev and better set the alarm timeout to a reasonable default or remove it at all.

Then please remove it.

Signature · 11 lines
Alexander E. Patrakov


-------------------------------------------------------
This SF.Net email is sponsored by:
Power Architecture Resource Center: Free content, downloads, discussions,
and more. http://solutions.newsforge.com/ibmarch.tmpl
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel