David Boles wrote:
You are going to have explain the logic for this one for me to
understand.
"Excessive fsync during a kernel compile causes Firefox 3 become
completely unresponsive till the fsyncs are complete. In some cases if
something else i/o intensive is going on Firefox 3 will freeze till the
other i/o has completely finished. If it gets really really bad other
applications start freezing."
You were doing something as intensive as compiling a kernel and Firefox
became "unresponsive" and that is Firefox's fault?
It is about perspective. From one perspective it is Firefox 3's fault,
in another it is the ext3 code's fault.
In the first perspective Firefox 2 and Firefox 3.0b2 don't do it, but
Firefox 3.0b3 and later do. In general I would say it might even be fine
for Firefox to do one fsync per page, but it seems to do like eight back
to back. Which in my book is excessive. Maybe there is a good reason for
it, maybe it can be improved. I don't claim to know much about Firefox
3's code, but I can give the evidence to the programmers and let them
sort it out.
In the second perspective it is more of a problem with ext3 code. It
doesn't handle fsyncs that well when it comes to responsiveness. I do
plan to take this up with the maintainers of the ext3 code. If I can't
get them to fix this to my satisfaction I will switch to another
filesystem, and maybe another distribution that supports that filesystem
well.
