PAE bug in 2.4?!
Hi,
I've ASUS P4P800-Mx-EAYVZ motherboard and have to use a 2GB memory. But after including PAE support in kernel it almost twice... Surely, i understand, that it's inevitable, the slowing of PAE, but twice?! Does anybody faced this problem? Or is there some bug in PAE?! I tested the PAE at some other motherboards, but only on this model on asus it works strange)
Best rgds, Ilya
Below, you can find some usefull info about this situation.
Cpu info:
khext@al:~$ cat /proc/cpuinfo processor : 0 vendor_id : GenuineIntel cpu family : 15 model : 2 model name : Intel(R) Celeron(R) CPU 2.40GHz stepping : 9 cpu MHz : 2400.156 cache size : 128 KB fdiv_bug : no hlt_bug : no f00f_bug : no coma_bug : no fpu : yes fpu_exception : yes cpuid level : 2 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe cid bogomips : 4784.12
Here is the difference in kernels:
khext@al:~$ diff linux-2.4.31/.config linux-2.4.31-PAE/.config 61,62c61,62 < CONFIG_NOHIGHMEM=y < # CONFIG_HIGHMEM4G is not set ---
Quoted 2 lines
#
CONFIG_NOHIGHMEMis not setCONFIG_HIGHMEM4G=y
64c64,65
< # CONFIG_HIGHMEM is not set
---
Quoted 2 lines
CONFIG_HIGHMEM=yCONFIG_HIGHIO=y
Here is my test:
khext@al:~$ cat pl
for ($j=1;$j<5;$j++){
$num=250000*$j;
$a="";
for ($i=0; $i<$num/72; $i++) {
$a.='1234567890qwertyuiopasdfghjklzxcvbnm';
$a.='1234567890QWERTYUIOPASDFGHJKLZXCVBNM';
}
printf ("len=".length($a)."\n");
$b="";
$k=345345342;
for ($i=0; $i<1000000; $i++) {
$k*=34523;
$k%=$num;
$k=abs($k);
$b.=substr(substr($a, $k, 4097),$k%4097,1);
}
printf ("b=".length($b)."\n");
printf (substr($b, 10, 5)."\n");
}Test themselfs...
With PAE turned on:
khext@al:~$ uname -a
Linux al 2.4.31 #25 SMP Wed Nov 2 19:05:47 MSK 2005 i686 GNU/Linux
khext@al:~$ free
total used free shared buffers
cached
Mem: 2059812 30656 2029156 0 2512
9744
-/+ buffers/cache: 18400 2041412
Swap: 4883720 0 4883720
khext@al:~$ time perl pl
len=250056
b=991920
xPnx7
len=500040
b=996000
SqnSb
len=750024
b=997280
e2ux7
len=1000008
b=997960
SLuSbreal 0m33.436s user 0m33.210s sys 0m0.220s
With PAE turned off:
khext@al:~$ uname -a
Linux al 2.4.31 #25 SMP Wed Nov 2 19:00:05 MSK 2005 i686 GNU/Linux
khext@al:~$ free
total used free shared buffers
cached
Mem: 903880 27640 876240 0 2492
9756
-/+ buffers/cache: 15392 888488
Swap: 4883720 0 4883720khext@al:~$ time perl pl len=250056 b=991920 xPnx7 len=500040 b=996000 SqnSb len=750024 b=997280 e2ux7 len=1000008 b=997960 SLuSb
real 0m17.208s user 0m16.780s sys 0m0.430s
Quoted 6 lines · On Wed, 02 Nov 2005 18:49:55 +0300 Ilya wrote:
I've ASUS P4P800-Mx-EAYVZ motherboard and have to use a 2GB memory. But after including PAE support in kernel it almost twice... Surely, i understand, that it's inevitable, the slowing of PAE, but twice?! Does anybody faced this problem? Or is there some bug in PAE?! I tested the PAE at some other motherboards, but only on this model on asus it works strange)
[skip]
Quoted 6 lines
Here is the difference in kernels:
khext@al:~$ diff linux-2.4.31/.config linux-2.4.31-PAE/.config 61,62c61,62 < CONFIG_NOHIGHMEM=y < # CONFIG_HIGHMEM4G is not set ---
Quoted 7 lines
Quoted 2 lines
#
CONFIG_NOHIGHMEMis not setCONFIG_HIGHMEM4G=y64c64,65
< #CONFIG_HIGHMEMis not set
---Quoted 2 lines
CONFIG_HIGHMEM=yCONFIG_HIGHIO=y
CONFIG_HIGHMEM4G does not enable PAE - it is CONFIG_HIGHMEM64G which does it. So the slowdown must be due to something other than PAE.
Please also show the start of dmesg output - lines which look like:
BIOS-provided physical RAM map:
BIOS-e820: 0000000000000000 - 000000000009fc00 (usable)
...
xxxMB HIGHMEM available.
xxxMB LOWMEM available.
Also show the output of "cat /proc/mtrr". Some systems have buggy MTRR setup done by BIOS, which can lead to such slowness when using the full amount of memory. Such problems may be fixed by a BIOS upgrade, or worked around by specifying the mem=... option to avoid using the memory which is not covered by MTTRs (if the amount of such misconfigured memory is not excessively large).