Lost hours of gameplay due to power cut - sstate is there but pcsx2 wont load it
#21
(08-28-2011, 06:49 PM)nosisab Ken Keleh Wrote: Indeed, I meant this is outside PCSX2 responsibility or scope even. It's not a bug, the disc cache is a compromise (OS concerned) and even so, not even disabling it can grant complete security, mainly on hard bsods (The UPS can grant the time to close all handlers before shutting down the system on power failure but can't do anything if the OS itself crashes).

If this helps your discussion - the power cut did not occur during saving the game, or even 2s after it. It happened at least a clean 4 minutes or so.
Reply

Sponsored links

#22
(08-28-2011, 10:49 PM)perkyguy Wrote: If this helps your discussion - the power cut did not occur during saving the game, or even 2s after it. It happened at least a clean 4 minutes or so.

Flushing the cached writes can take much more than 2s (on a clean windows XP SP3 it can take upto 15-25s, tested). I have no info about the longest it might take, but I'd consider 4 mins way too much, but not impossible, especially if the CPU is at high loads (I'd imagine the cache writes are lower priority, as BSODS/power-off happen very rarely).

That being said, if the issue can't be reproduced, it's going to be extremely hard to fix, and so far, I could not find any evidence that this is indeed a pcsx2 bug. I'm not saying that your numbers are incorrect or that a bug doesn't exist, but at the lack of better info, I'd have to still consider it a delayed cache write.

If anyone is able to reproduce the issue with reasonable repeatability, by all means let us know and we'll look into it again. For now though, I'm dismissing these issues as cached writes data-loss on power-off.
Reply
#23
(08-28-2011, 05:48 PM)nosisab Ken Keleh Wrote: A more pausible cause of (presumed saved) data loss (of any type) on power failure is the disc cache feature, a sad compromise I don't know a way to solve other than using UPS (Uninterruptible Power Supply) or disabling the cache which would mean baaad news to performance.

eh, not really, there are functions to ensure a file is fully written to disk and ensure its not living in disk cache.

avih, theres not much you can really do to give a full repro on it except simulate the event, which i have no intention of following through on with my important data.

watching the app writes with diskmon and the threads with process monitor seems to point to the file handle remaining open for longer then it should in some cases, which is enough to enact an investigation.

I would start with updating the zlib library though, because the version used does has an issue with threads stalling and hanging the parent application.
Reply
#24
(08-29-2011, 05:02 AM)Squall Leonhart Wrote: ...
watching the app writes with diskmon and the threads with process monitor seems to point to the file handle remaining open for longer then it should in some cases, which is enough to enact an investigation.
...

Could you share, in detail, your findings/data/etc regarding the file handle? my tests with processMon suggests it's always closed on time.
Reply
#25
how long were you playing for before checking?

of course, it works fine if you only have pcsx2 open for 10ish minutes, but some of us are playing for hours on end using the states system to save/reload countless times where it benefits us.

im seeing the file handle stay open for almost 2 minutes after 6 hours of play, and i don't know (it might) if running with the fps uncapped exacerbates the problem.

the zlib dev is encouraging developers to upgrade to 1.2.5 immediately however, due to some issues present in 1.2.4,
Reply
#26
(08-29-2011, 09:19 AM)Squall Leonhart Wrote: ...
im seeing the file handle stay open for almost 2 minutes after 6 hours of play, and i don't know (it might) if running with the fps uncapped exacerbates the problem.
...

Is that after the console message "(gzipThread) Data saved to disk without error."? Or does the message itself appear minutes after the save was initiated?

Also, could you describe the procedure (and tools) in detail?
Reply
#27
Well, I assumed the problem arose from the power failure occurring soon after the save, those minutes span really run out disc cache as guilt. As it turns out jumping over conclusions is fairly risky, maybe the better is the OP sending a copy of that sstate to be analyzed. Until them is better not to assume is a PCSX2 problem/bug ever... Power failure is somewhat like natural disasters...

To the OP is advised to run chkdsk on all his/her discs/partitions, not because this particular issue but as good measure after power failure in general.

Imagination is where we are truly real
Reply




Users browsing this thread: 1 Guest(s)