Re: ps performance sucks (was Re: dcache_rcu [performance results])

[Kai Henningsen]

[e-mail] (Martin J. Bligh)  wrote on 04.11.02 in <1118170000.1036458859@flay>:

> I had a very brief think about this at the weekend, seeing
> if I could make a big melting pot /proc/psinfo file that did
> seqfile and read everything out in one go, using seq_file
> internally to interate over the tasklist. The most obvious
> problem that sprung to mind seems to be the tasklist locking -
> you obviously can't just hold a lock over the whole thing.

Well, one thing i to make certain you can actually do it with one or two  
system calls. Say, one system call to figure out how big a buffer is  
necessary (essentially, #tasks*size), then one read with a suitably-sized  
buffer. Then have a loop in the kernel that drops the lock as often as  
necessary, and otherwise puts it all in the buffer in one go. (If the  
#tasks grows too fast so it overruns the buffer even with some slack given  
in advance, tough, have a useful return code to indicate that and let ps  
retry.)

I briefly thought about mmap, but I don't think that actually buys  
anything.

MfG Kai