Skip to contentSkip to conversation

PCI oddities on Dell Inspiron 5000e w/ 2.4.x

DISCUSSION Jeff Lessem 4 participants 5 messages Feb 24, 2001

lore
Share & tools
lore
#0 lore raw

In your message of: Fri, 23 Feb 2001 09:48:04 PST, you write:

Quoted 11 lines

The much more likely cause is the "magic registers" for the Texas Instruments PCI1225, namely

		works		broken

	81:	b0		90
	a8:	11		10

Although it worries me a bit that your second controller also seems to have differences in the BridgeCtl thing (16bInt).

Quoted 6 lines

Can you try if a broken setup is fixed by doing a

	setpci -s 00.04.0 81.b=b0
	setpci -s 00.04.0 a8.b=11
	setpci -s 00.04.1 81.b=b0
	setpci -s 00.04.1 81.b=11

I ran setpci -s 00:04.0 81.b=b0, etc., but it didn't make any difference. Checking with lspci -vvxxx after running setpci it appears that register 81 and a8 have stayed at 90 and 10 despite setpci running without an error, using -G and -v setpci claims to be running through /proc/bus/pci and adjusting the appropriate location. Either I am making a fundamental error (yes, I am running it as root) or the changes simply don't matter.

Quoted 2 lines

Also, how much memory does this machine have? That "13ff0000" does worry me a bit..

The comptuer has 320MB. At this point I am ready to conclude that the computer is broken in some way, because nobody else with an Inspiron 5000e that I have heard from has anything like this problem.

I really appreciate the help everybody is providing with what is really only an annoyance, and I wouldn't have even brought it to the attention of linux-kernel, except that something that changed between 2.2.17 and 2.4.x has produced a regression in functionality. If anybody thinks it would be useful, I would be willing to do a binary search between 2.2.17 and 2.4.2 to find out at which release things stopped being setup correctly.

--
Jeff Lessem.

#1 lore raw
Quoted 7 lines · Jeff Lessem
Quoted 3 lines

Also, how much memory does this machine have? That "13ff0000" does worry me a bit..

The comptuer has 320MB. At this point I am ready to conclude that the computer is broken in some way, because nobody else with an Inspiron 5000e that I have heard from has anything like this problem.

The machine isn't broken. It's Linux.

I didn't believe that you'd have 320MB just because it's such an odd number, but the problem is that Linux apparently starts allocating the PCI address space just _under_ the 320MB mark (you probably have some memory reserved for ACPI that doesn't show up in the e820 memory map).

The PCI allocation needs to be fixed, but I'm off for two days in Finland..

Linus

#2 lore raw
In reply to Jeff Lessem
Quoted 5 lines · Linus Torvalds
Quoted 4 lines · Jeff Lessem
Quoted 3 lines

Also, how much memory does this machine have? That "13ff0000" does worry me a bit..

Quoted 8 lines
Quoted 3 lines

The comptuer has 320MB. At this point I am ready to conclude that the computer is broken in some way, because nobody else with an Inspiron 5000e that I have heard from has anything like this problem.

I didn't believe that you'd have 320MB just because it's such an odd number, but the problem is that Linux apparently starts allocating the PCI address space just _under_ the 320MB mark (you probably have some memory reserved for ACPI that doesn't show up in the e820 memory map).

Jeff, are you using the e820 memory map at all ? In particular, are you using grub or some other buggy bootloader that insists on specifying a mem= option on the kernel command line ? There should be a kernel command line message very early on, what does that say ?

Also, can you give us the E820 memory map (kernel messages starting with BIOS-e820) ?

#3 lore raw
In reply to Jeff Lessem
Quoted 23 lines · Jeff Lessem

In your message of: Sat, 24 Feb 2001 09:54:47 CST, you write:

Quoted 4 lines

Jeff, are you using the e820 memory map at all ? In particular, are you using grub or some other buggy bootloader that insists on specifying a mem= option on the kernel command line ? There should be a kernel command line message very early on, what does that say ?

Yes, I am using grub, the buggy bootloader. The relevant chunk of kernal messages are:

BIOS-provided physical RAM map:
BIOS-e820: 000000000009f800 @ 0000000000000000 (usable)
BIOS-e820: 0000000000000800 @ 000000000009f800 (reserved)
BIOS-e820: 0000000000019800 @ 00000000000e6800 (reserved)
BIOS-e820: 0000000013ef0000 @ 0000000000100000 (usable)
BIOS-e820: 000000000000fc00 @ 0000000013ff0000 (ACPI data)
BIOS-e820: 0000000000000400 @ 0000000013fffc00 (ACPI NVS)
BIOS-e820: 0000000000080000 @ 00000000fff80000 (reserved)
On node 0 totalpages: 81904
zone(0): 4096 pages.
zone(1): 77808 pages.
zone(2): 0 pages.
Kernel command line: root=/dev/hda1 mem=327616K

Quoted 2 lines

You are dead on, mem= seems a bit small. Forcing mem=320M on the command line fixes the problem completely.

Careful, you're overwriting ACPI data now (and using it as normal RAM).

Can you try one of a) LILO b) a fixed version of grub c) this patch ?

arch/i386/kernel/setup.c +2 -23
diff -ur linux/arch/i386/kernel/setup.c linux-prumpf/arch/i386/kernel/setup.c
@@ -555,30 +555,9 @@
 				e820.nr_map = 0;
 				usermem = 1;
 			} else {
-				/* If the user specifies memory size, we
-				 * blow away any automatically generated
-				 * size
-				 */
-				unsigned long start_at, mem_size;
- 
-				if (usermem == 0) {
-					/* first time in: zap the whitelist
-					 * and reinitialize it with the
-					 * standard low-memory region.
-					 */
-					e820.nr_map = 0;
-					usermem = 1;
-					add_memory_region(0, LOWMEMSIZE(), E820_RAM);
-				}
-				mem_size = memparse(from+4, &from);
+				memparse(from+4, &from);
 				if (*from == '@')
-					start_at = memparse(from+1, &from);
-				else {
-					start_at = HIGH_MEMORY;
-					mem_size -= HIGH_MEMORY;
-					usermem=0;
-				}
-				add_memory_region(start_at, mem_size, E820_RAM);
+					memparse(from+1, &from);
 			}
 		}
 		c = *(from++);
#4 lore raw
In reply to Jeff Lessem
Quoted 4 lines

In your message of: Sat, 24 Feb 2001 11:55:07 CST, you write:

Quoted 2 lines

Careful, you're overwriting ACPI data now (and using it as normal RAM).

Quoted 8 lines

Hmm, I guess that would be bad.

Quoted 1 line

Can you try one of a) LILO b) a fixed version of grub c) this patch ?

I tried LILO and the problem did indeed go away when using that. I guess I'll stick with LILO until Linux or grub (whichever is broken) is fixed. There is just something appealing about a proper boot console on a PC...

Even though I wasn't much help on this, it's nice to know what was different on your config that was causing problems while I wasn't seeing any issues. I sat here for almost an hour last night trying to figure out why your machines memory map would different than mine (I have 320MB of RAM as well, so they should have been the same). The thought that you might be using a different boot loader never even crossed my mind. I've always used LILO so that's why I didn't see any problems.

It's nice to know it wasn't just working for me by accident.

Later,
Tom