Ticket #179 (closed bug: duplicate)

Opened 6 years ago

Last modified 6 years ago

ffado 2.0 and swap creates xruns

Reported by: dx9s Assigned to:
Priority: major Milestone: FFADO 2.0
Component: Version: FFADO SVN (2.0 branch)
Keywords: b1484 Cc:
The device the bug applies to:

Description

I **stumbled** on this by accident (short/long story about disabling swap and reusing that partition to try to install XP and flash firmware)

When I was using my Echo AF12 initially it worked flawlessly (and I had no swap partition enabled) ... solid for 20 minutes at 3x64 and 48kHz -- even storing WAVs to external USB2 HD.

At some point after that I wiped the XP partition (never used it for updating firmware, brought AF12 to work and [ab]-used an XP+SP2 machine there) and re-enabled swap at boot (/etc/fstab)

and I was getting XRUNS regardless of what jackd settings used with ffado. Just COULDN'T stop them, and I was smart to remember the change I did and reversed it (swapoff /dev/sdX) and restarted and tada SOLID

so a work around for those having XRUNs and for testing purposes is to disable swap memory (can script it inside qjackctl via gksudo swapoff /dev/sdX)

I have a sneaky suspicion that some buffer and/or code space (perhaps all of it) isn't properly "locking" it's memory or something like that. Since I don't fully understand the details... I'll shut-up and leave it at that.

--Doug

Change History

11/27/08 09:46:50 changed by dx9s

A "feature" that would be nice (should open up a new ticket if this sounds smart) is a dummy ffado streamer mode (no actual hardware). I do benchmarks against jackd in dummy mode to test it's stability -- would make sense to have a flag when starting ffado-dbus-server to cause the streamer part to go into a multi-channel loopback mode. And test for stability there before moving onto actual hardware.... just a thought.

--Doug

11/27/08 12:34:13 changed by dx9s

Ubuntu uses "sw" for defaults in fstab .. I changed to "defaults" and now I can't get the machine to pre-produce the xruns at reasonable jackd settings... system is stable at 3x64 w/ no swap or with swap (except for ACPI events like power<->battery state changes) .. 3x128 is long enough to cover battery state changes in acpid ...

back to fstab... when it was "sw" for options in fstab ... there was NO option I could set jackd at and NOT get an xrun in one minute (or less)...

This is getting stranger by the moment. If I can get it to reproduce reliably -- I will update here and provide that to you'all ... Until then!

--Doug

11/27/08 12:58:33 changed by dx9s

well my test outside ardour was not good enough ... testing inside ardour really is what counts and swap does effect some.. and still testing ... I can get xruns but it's not like it was before (about one-two a minute at what seemed regular intervals) -- that's gone / now I think I'm experiencing pure hardware performance.... I haven't switched back to "sw" in options yet ... so no telling have lots of testing to do.

11/27/08 13:32:57 changed by kooox

AFAIK, the "sw" option in fstab is just used to mark which devices swapon/off can enable/disable when launched with the -a option. This could simply mean that when a swap partition does not have "sw" in /etc/fstab, it is _not_ activated on boot. I cannot verify that right now, but you may want to check if your swap is activated when "sw" is not present (e.g. with cat /proc/swaps). If it's not, then not having "sw" in /etc/fstab is just like not having a swap partition at all (except that in the latter case, you cannot even activate it manually, of course).

11/28/08 10:46:49 changed by dx9s

  • status changed from new to closed.
  • resolution set to duplicate.

I am closing this in place of ticket 181 as ffado has issues with SMP it appears and is exposed under strange situations. I am marking this closed/duplicate see 181 as that better describes the variables and test I have performed.