Serious problem with multithreaded I/O on Linux

[Tom Tromey]

>>>>> "Bryce" == Bryce McKinlay <bryce@albatross.co.nz> writes:

Bryce> Hmm... This sounds like a nasty problem. If boehm is making I/O
Bryce> calls "gc-safe" by blocking all threads when they are used, how
Bryce> is combining them going to improve the situation?

Bryce> Why are blocking IO calls not gc-safe, anyway?

I don't think it wants to block all the threads (though I could be
wrong as I haven't dug through all that code yet).  Instead I believe
it is changing the protection on the page so that the read() works.
In incremental mode the Boehm GC write-protects pages to implement a
write barrier.

Tom