Messages in this thread | | | Date | Fri, 4 Jan 2008 16:49:06 -0800 (PST) | From | Linus Torvalds <> | Subject | Re: [patch] slub: provide /proc/slabinfo |
| |
On Fri, 4 Jan 2008, Arjan van de Ven wrote: > > this is the part I'm not very thrilled about... at least on first sight > it looks like a user can now hold the read sem over system calls, and > for as long as it wants.
No, you misunderstand how seq-files work.
The start/stop sequence is done only overone single kernel buffer instance, not over the whole open/close (or even the copy to user space).
It's expressly designed so that you can hold locks (including spinlocks etc), and will not do any blocking ops (well - your *callbacks* can you blocking ops, but the seq_file stuff itself won't) between start/stop.
Linus
| |