On Thu, 24 May 2001, Jaswinder Singh wrote:
> Dear Mike ,
>
> >
> > This one I tested with memleak. It wasn't a leak, it was dcache
> > growth. Under vm stress, it shrank down fine.
> >
>
> It will depends upon lot of thing :-
> 1.What is your size of ramfs ,
unlimited.
> 2. Are you using any harddisk ,
(?? ramfs) I think I misunderstand.
> 3. How many many files are you creating ,
lots of different names. (like you were)
> 4. How frequently you are making files .
as fast as a C proglet could create/maybe(random)access/destroy them.
> 5. What are the size of your files ?
varying from empty to small.
>
> In my case , my ramfs is of 14 MB , using no hard-disk , i am continously
> making and deleting files of 4 K size , my Machine hangs within 5 minutes ,
> by saying no memory left.
I can write random 4k files as fast and long as I like (with swap off)
and if I flush out the rather large dcache...
procs memory swap io system cpu
r b w swpd free buff cache si so bi bo in cs us sy id
2 0 1 0 17600 104 4456 0 0 197 4 118 41 6 83 11
1 0 0 0 14888 124 4476 0 0 31 0 123 37 13 87 0
1 0 0 0 12252 124 4476 0 0 0 0 103 9 11 89 0
1 0 0 0 9632 124 4476 0 0 0 0 105 7 17 83 0
1 0 0 0 7032 124 4480 0 0 0 0 103 10 11 89 0
1 0 0 0 4460 124 4476 0 0 0 0 101 10 12 88 0
1 0 1 0 20272 128 4480 0 0 1 2 116 16 23 77 0
1 0 0 0 17860 128 4480 0 0 0 15 115 10 17 83 0
1 0 0 0 15196 128 4476 0 0 0 0 110 7 15 85 0
dentry_cache 747962 747990 128 24933 24933 1
dentry_cache 769651 769680 128 25656 25656 1
dentry_cache 792664 792690 128 26423 26423 1
dentry_cache 814779 814800 128 27160 27160 1
dentry_cache 835972 835980 128 27866 27866 1
dentry_cache 715148 715170 128 23839 23839 1
...afterward, I don't see leakage. Only a short test this time, but
long enough to eat all of my ram many times over if it leaked as fast
as your description indicates. I can't reproduce it. (2.4.4 btw)
> Please tell me about your system and what you are doing on it ?
Generic 128mb PIII/500. What I was doing on it was beating up ramfs
looking for the leak you reported :)
-Mike
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Thu May 31 2001 - 21:00:22 EST