Koozali.org: home of the SME Server
Legacy Forums => General Discussion (Legacy) => Topic started by: geoff on November 28, 2002, 05:02:22 AM
-
Hi, All ...
I am setting up a Server as follows:
Pentium4 2GhxCPU, 512MbRAM, MSI6398E-LR Mainboard, Adaptec 29160N SCSI Controller with 2 X Atlas 36Gb SCSI HDD which I want to software mirror.
The load is perfect, then upon reboot it gets to where devices are probed and speed established (8 regs), then it probes for SCSI. It is looking for Channel 0, whereas the SCSI controller is set for Channel A.
I'm baffled now. Can anyone help, please?
-
Further, the following is displayed:
Detected scsi disk sda0 at scsi0 channel 0 id 0 lun 0
scsi: aborting command due to timeout ....
This continues for sdb and so on without resolution. The SCSI adapter seems to be set to Channel A although e-smith seems ot be looking for Channel 0.
Very puzzled .....
-
geoff wrote:
> Detected scsi disk sda0 at scsi0 channel 0 id 0 lun 0
> scsi: aborting command due to timeout ....
Such things are often due to cabling or termination issues.
> The SCSI
> adapter seems to be set to Channel A although e-smith seems
> ot be looking for Channel 0.
What's in a name? Probably the same thing.
Charlie
-
Thank you for your reply, Charlie, but I have been there and checked termination (the terminator is OK), and the two drives are plugged to the last two connectors on the cable. I have also set 'Termination' in the SCSI BIOS to 'Enabled' rather than 'Automatic'. I just loaded Win2000 Server with software RAID successfully but no success with e-smith 5.5, 5.1.2, or 4.1.2.
It is very like the issues experienced with Adaptec and Novell a few years ago where the driver was critical. Further, I have just been told that the 29160N is a Workstation adapter that sometimes causes issues with Server installations. Weird.
I am waiting for a 29160 (no suffix) to try. Also we have an IBM e-Server running a 29160LP with no problems at all.
Again, thank you for your reply.
-
Just tried it with a 29160 Adapter and it acted in exactly the same manner.
Then tried startup and selected 'Mitel UP' (instead of the default 'Mitel') and everything is fine with either the 29160N or 29160 SCSI Controllers ....... mutter, mutter, mutter ...
Question: Why is a single processor Mainboard being detected as multi-processor?
-
geoff wrote:
> Question: Why is a single processor Mainboard being detected
> as multi-processor?
Here is the detection code used by the RedHat anaconda installer:
...
/* MP Floating Pointer Structure */
typedef struct MPFPS {
char signature[ 4 ];
void* pap;
u_char length;
u_char spec_rev;
u_char checksum;
u_char mpfb1;
u_char mpfb2;
u_char mpfb3;
u_char mpfb4;
u_char mpfb5;
} mpfps_t;
...
static int intelDetectSMP(void)
{
vm_offset_t paddr;
int where;
mpfps_t mpfps;
int rc = 0;
int ncpus = 0;
/* open physical memory for access to MP structures */
if ( (pfd = open( "/dev/mem", O_RDONLY )) < 0 ) {
return 0;
}
/* probe for MP structures */
apic_probe( &paddr, &where );
if ( where <= 0 )
return 0;
if (seekEntry( paddr ))
return 0;
readEntry( &mpfps, sizeof( mpfps_t ) );
if (mpfps.mpfb1)
/* old style */
rc = 1;
else {
/* go to the config table */
mpcth_t cth;
int count, i;
paddr = (vm_offset_t) mpfps.pap;
if (seekEntry( paddr ))
return 0;
readEntry( &cth, sizeof( cth ) );
/* if we don't have any entries, the kernel sure
won't be able to set up mp. Needs at least one entry
for smp kernel */
if (cth.entry_count <= 1) {
close (pfd);
return 0;
}
count = cth.entry_count;
for (i = 0; i < count; i++) {
if ( readType() == 0 ) {
ProcEntry entry;
readEntry( &entry, sizeof( entry ) );
if (entry.cpuFlags & PROCENTRY_FLAG_EN)
ncpus++;
}
}
if (ncpus > 1)
rc = 1;
}
close (pfd);
return rc;
}
This says that you get an SMP kernel if you have more than one CPU, or you have a motherboard with a valid MultiProcessor Floating point structure, with non-zero mpfb1 element.
Charlie
-
ah i reckon iv eseen this !
its the motherboard chip set
(and is reported somewhere in the forum)
that causes the default load to be the smp
funny thing, the first time i used sme (5.5),
i had this problem, so i kludged the /etc/lilo.conf
to correct the default and all ok -
i just assumed (being the 1st time) sme always did this LOL
regards Rob
-
Thanks to everyone for their help.
-
Hi Robert,
great for people, who have only 1 CPU!
i have several servers with 2 or 4 XEONs and without SME 5.6 and the newer kernel, the rest of them are still useless....I can use them only in UP-Mode...sniff...
Greeting Michael