Re: [PATCH] fs/devfs/base.c

[Richard Gooch <rgooch@ras.ucalgary.ca>]

Alan Cox writes:
> >  - the kernel stack is 4kB, and _nobody_ has the right to eat up a
> >    noticeable portion of it. It doesn't matter if you "know" your caller
> 
> Umm Linus on what platform - its 8K or more on all that I can think of

I assume he's referring to what's left after you take out the task
struct. More than 4 kiB, so 4 kiB is a conservative estimate.

> > Ergo: the simple rule of "don't allocate big structures of the stack" is
> > always a good rule, and making excuses for it is bad.
> 
> We have a very large number of violators of 1K of stack, and very
> few of 2K right now.

I guess we should ask the question as to what's an acceptable usage.
Theoretically, any amount could pose a problem, but that's hardly a
useful position to work from. Some decision or guideline about a
practical limit would be helpful. My gut feeling is that 1 kiB is
around the limit for functions with a well defined call path. The
limit should probably be less for generic functions and interrupt
handlers.

				Regards,

					Richard....
Permanent: rgooch@atnf.csiro.au
Current:   rgooch@ras.ucalgary.ca