Re: BFS vs Tuxonice issue

Hi.

On 24/04/12 04:08, Harald Judt wrote:
> Hi Nigel,
>
> Did you have time to look into the BFS compatibility issue when 
> multi-threaded compression is enabled?
>
I got some debugging done, managing to confirm that it's definitely 
something going wrong when the image is written (I can enable 
multithreaded I/O for reading only and resume successfully). I also 
confirmed that there's no problem in the compressing itself (even 
running multithreaded, I can compress a page and decompress the result 
straight away without a problem). So I'm almost certain it's something 
in the paths used to coalesce the compressed pages and write them to 
disk. The hard part is figuring out what, especially when there's a 
mutex and a spinlock protecting the two parts of the code path that 
would be most likely to cause the problem!

I'll keep looking.

Nigel