Skip to contentSkip to conversation

make X server on zx2000 (and probably others) work again

PATCH David Mosberger 2 participants 3 messages Oct 22, 2003 EFI

lore
Share & tools
lore
#0 lore raw

In recent 2.5/6 kernels, the X server on my zx2000/Radeon workstation would cause a crash. Turns out this is due to a long-standing bug in efi.c: the range checking in efi_mem_type() and efi_mem_attributes() could cause an underflow if the size of a memory descriptor is zero (as happens sometimes due to memory trimming). In my case, granule zero had a zero size and that in turn caused all /dev/mem mappings to occur with WRITE-BACK mapping, which doesn't make the X server happy (instant MCA on a zx2000, for example).

Note that this bug existed for a long time, so it's possible that there may have been other subtle memory-attribute related problems in the past. This bug just became more noticable recently, since the /dev/mem driver now completely ignores the O_SYNC flag and relies totally on the value returned by efi_mem_attributes().

	--david

# This is a BitKeeper generated patch for the following project:
# Project Name: Linux kernel tree
# This patch format is intended for GNU patch command version 2.5 or higher. # This patch includes the following deltas:
# ChangeSet 1.1354 -> 1.1355
# arch/ia64/kernel/efi.c 1.25 -> 1.26
#
# The following is the BitKeeper ChangeSet Log
# --------------------------------------------
# 03/10/21 [email protected] 1.1355
# ia64: Fix efi_mem_type() and efi_mem_attributes() to avoid potential
# underflows. In my case, the underflows occurred with the
# first memory descriptor which got trimmed down to a size of 0.
# Due to the underflow, this descriptor ended up covering the entire
# address-range which in turn caused Bad Things to happen with the
# X server.
# --------------------------------------------
#

arch/ia64/kernel/efi.c +2 -2
diff -Nru a/arch/ia64/kernel/efi.c b/arch/ia64/kernel/efi.c
@@ -685,8 +685,7 @@
 	for (p = efi_map_start; p < efi_map_end; p += efi_desc_size) {
 		md = p;
 
-		if ((md->phys_addr <= phys_addr) && (phys_addr <-		    (md->phys_addr + (md->num_pages << EFI_PAGE_SHIFT) - 1)))
+		if (phys_addr - md->phys_addr < (md->num_pages << EFI_PAGE_SHIFT))
 			 return md->type;
 	}
 	return 0;
@@ -706,8 +705,7 @@
 	for (p = efi_map_start; p < efi_map_end; p += efi_desc_size) {
 		md = p;
 
-		if ((md->phys_addr <= phys_addr) && (phys_addr <-		    (md->phys_addr + (md->num_pages << EFI_PAGE_SHIFT) - 1)))
+		if (phys_addr - md->phys_addr < (md->num_pages << EFI_PAGE_SHIFT))
 			return md->attribute;
 	}
 	return 0;
#1 lore raw
Quoted 9 lines · David Mosberger

In recent 2.5/6 kernels, the X server on my zx2000/Radeon workstation would cause a crash. Turns out this is due to a long-standing bug in efi.c: the range checking in efi_mem_type() and efi_mem_attributes() could cause an underflow if the size of a memory descriptor is zero (as happens sometimes due to memory trimming). In my case, granule zero had a zero size and that in turn caused all /dev/mem mappings to occur with WRITE-BACK mapping, which doesn't make the X server happy (instant MCA on a zx2000, for example).

Quoted 6 lines

Note that this bug existed for a long time, so it's possible that there may have been other subtle memory-attribute related problems in the past. This bug just became more noticable recently, since the /dev/mem driver now completely ignores the O_SYNC flag and relies totally on the value returned by efi_mem_attributes().

I have a very strange problem with my bigsur. Whenever I put a new kernel on it, it would crash at random without any warnings during boot after a warm reboot. I had to power cycle it to get it to boot. After applying this patch to 2.4.21, I have rebooted it twice without powering it down first.

H.J.

#2 lore raw
In reply to David Mosberger
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line · H. J. Lu

HJ> I have a very strange problem with my bigsur. Whenever I put a HJ> new kernel on it, it would crash at random without any warnings HJ> during boot after a warm reboot. I had to power cycle it to get HJ> it to boot. After applying this patch to 2.4.21, I have rebooted HJ> it twice without powering it down first.

Interesting. Let's hope it has fixed this issue for good!

	--david