After I did some testing, I must admit that you did excellent work
Bill. Well, as always. :)
I applied the patch on 2.1.55 (!) and after being satisfied with
results, decided to benchmark things, just to be sure everything is
OK.
Here it comes (very simple):
(reboot to single user, login, cd /usr/src/linux)
free
time ls -alR / > /dev/null (2x to see caching effect)
free
time make -j10
free
(make clean; reboot)
make is done on Linux kernel (of course :) ) with MAKE=make -j10 in
Makefile.
Results:
2.1.55
______
total used free shared buffers cached
Mem: 30664 4908 25756 2064 900 2784
-/+ buffers: 1224 29440
Swap: 104796 0 104796
01:53 -> 00:32 (ls -alR real times)
total used free shared buffers cached
Mem: 30664 30084 580 2092 7476 2816
-/+ buffers: 19792 10872
Swap: 104796 0 104796
14:25 minutes to build kernel (make -j10)
total used free shared buffers cached
Mem: 30664 19832 10832 760 1480 6088
-/+ buffers: 12264 18400 <--- MEMORY LEAK!
Swap: 104796 224 104572
2.1.55 with inode patch applied (Bill Hawes)
____________________________________________
total used free shared buffers cached
Mem: 30664 4584 26080 2096 900 2476
-/+ buffers: 1208 29456
Swap: 104796 0 104796
01:32 -> 00:28 (some improvements even here!!!)
total used free shared buffers cached
Mem: 30664 14328 16336 2116 9924 1484
-/+ buffers: 2920 27744 <--- memory leak not present!!!
Swap: 104796 0 104796
08:56 minutes to build kernel!!!!!!!!!!!! (we saved 38% of time)
total used free shared buffers cached
Mem: 30664 9660 21004 760 1416 6064
-/+ buffers: 2180 28484
Swap: 104796 228 104568
Now I see 2.1.56 will be the best kernel so far!!!
Like other testers, I also got many "got one" messages, and not even
one "failed" message.
Excellent!
-- Posted by Zlatko Calusic E-mail: <Zlatko.Calusic@CARNet.hr> --------------------------------------------------------------------- Abandon all hope ye who have entered cyberspace.