Skip to contentSkip to conversation

Document hadling of bad memory

PATCH Pavel Machek 6 participants 12 messages Nov 26, 2008 – Dec 9, 2008 Documentation

lore
Share & tools
lore
#0 lore raw

Document how to deal with bad memory reported with memtest.

Documentation/bad_memory.txt new file +45 -0
new file mode 100644
index 0000000..df84162
@@ -0,0 +1,45 @@
+March 2008
+Jan-Simon Moeller, [email protected]
+
+
+How to deal with bad memory e.g. reported by memtest86+ ?
+#########################################################
+
+There are three possibilities I know of:
+
+1) Reinsert/swap the memory modules
+
+2) Buy new modules (best!) or try to exchange the memory
+   if you have spare-parts
+
+3) Use BadRAM or memmap
+
+This Howto is about number 3) .
+
+
+BadRAM
+######
+BadRAM is the actively developed and available as kernel-patch
+here:  http://rick.vanrein.org/linux/badram/
+
+For more details see the BadRAM documentation.
+
+memmap
+######
+
+memmap is already in the kernel and usable as kernel-parameter at
+boot-time.  Its syntax is slightly strange and you may need to
+calculate the values by yourself!
+
+Syntax to exclude a memory area (see kernel-parameters.txt for details):
+memmap=<size>$<address>
+
+Example: memtest86+ reported here errors at address 0x18691458, 0x18698424 and
+         some others. All had 0x1869xxxx in common, so I chose a pattern of
+         0x18690000,0xffff0000.
+
+With the numbers of the example above:
+memmap=64K$0x18690000
+ or
+memmap=0x10000$0x18690000
+
Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
#2 lore raw
In reply to Pavel Machek

[ [email protected] added, these should be the proper guys to
merge this ]

Quoted 61 lines · Pavel Machek

Document how to deal with bad memory reported with memtest.

Documentation/bad_memory.txt new file +45 -0
new file mode 100644
index 0000000..df84162
@@ -0,0 +1,45 @@
+March 2008
+Jan-Simon Moeller, [email protected]
+
+
+How to deal with bad memory e.g. reported by memtest86+ ?
+#########################################################
+
+There are three possibilities I know of:
+
+1) Reinsert/swap the memory modules
+
+2) Buy new modules (best!) or try to exchange the memory
+   if you have spare-parts
+
+3) Use BadRAM or memmap
+
+This Howto is about number 3) .
+
+
+BadRAM
+######
+BadRAM is the actively developed and available as kernel-patch
+here:  http://rick.vanrein.org/linux/badram/
+
+For more details see the BadRAM documentation.
+
+memmap
+######
+
+memmap is already in the kernel and usable as kernel-parameter at
+boot-time.  Its syntax is slightly strange and you may need to
+calculate the values by yourself!
+
+Syntax to exclude a memory area (see kernel-parameters.txt for details):
+memmap=<size>$<address>
+
+Example: memtest86+ reported here errors at address 0x18691458, 0x18698424 and
+         some others. All had 0x1869xxxx in common, so I chose a pattern of
+         0x18690000,0xffff0000.
+
+With the numbers of the example above:
+memmap=64K$0x18690000
+ or
+memmap=0x10000$0x18690000
+
Signature · 3 lines
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
Signature · 2 lines
Jiri Kosina
SUSE Labs
#3 lore raw
In reply to Pavel Machek
Quoted 1 line · On Wednesday 26 November 2008 10:15:21 Pavel Machek wrote:

Document how to deal with bad memory reported with memtest.

...

Documentation/bad_memory.txt line 23

Quoted 4 lines
Documentation/bad_memory.txt new file +4 -0
+BadRAM
+######
+BadRAM is the actively developed and available as kernel-patch
+here:  http://rick.vanrein.org/linux/badram/

So the patch isn't worth merging, but documentation about the out-of-tree patch is worth merging?

I'm not objecting, I'm just confused about to what the merge criteria are...

Rob

#4 lore raw

Am Freitag 28 November 2008 10:00:26 schrieb Rob Landley:

Quoted 3 lines

So the patch isn't worth merging, but documentation about the out-of-tree patch is worth merging?

Good point.

IIRC we tried merging the patch, but without luck at that time. It was said, that there's another method (with an even <irony>better</irony> syntax) which could also handle this case and there should be better some hacking to get the syntax parsed to use the functions of this already in-kernel method. I don't know the status of this (guess: none).
What I know: badmem worked here really good. (But meantime I bought new ram.)

Best regards,
Jan-Simon

#5 lore raw
In reply to Rob Landley
Quoted 10 lines · Rob Landley
Quoted 1 line · On Wednesday 26 November 2008 10:15:21 Pavel Machek wrote:

Document how to deal with bad memory reported with memtest.

...

Documentation/bad_memory.txt line 23

Quoted 4 lines
Documentation/bad_memory.txt new file +4 -0
+BadRAM
+######
+BadRAM is the actively developed and available as kernel-patch
+here:  http://rick.vanrein.org/linux/badram/

So the patch isn't worth merging, but documentation about the out-of-tree patch is worth merging?

Well, why not. The patch is unneccessary, but for the poor souls hit by bad memory, one line pointer can help...
Pavel

Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
#6 lore raw
Quoted 7 lines · On Friday 28 November 2008 06:18:38 Pavel Machek wrote:
Quoted 2 lines · Rob Landley

So the patch isn't worth merging, but documentation about the out-of-tree patch is worth merging?

Well, why not. The patch is unneccessary, but for the poor souls hit by bad memory, one line pointer can help...
Pavel

Define "unnecessary".

Rob

#7 lore raw
In reply to Rob Landley
Quoted 11 lines · On Fri, 28 Nov 2008 03:00:26 -0600 Rob Landley <[email protected]> wrote:
Quoted 1 line · On Wednesday 26 November 2008 10:15:21 Pavel Machek wrote:

Document how to deal with bad memory reported with memtest.

...

Documentation/bad_memory.txt line 23

Quoted 4 lines
Documentation/bad_memory.txt new file +4 -0
+BadRAM
+######
+BadRAM is the actively developed and available as kernel-patch
+here:  http://rick.vanrein.org/linux/badram/

So the patch isn't worth merging, but documentation about the out-of-tree patch is worth merging?

Quoted 2 lines

I'm not objecting, I'm just confused about to what the merge criteria are...

mm.. If someone finds it useful (and I assume that at least one person would have found it useful, hence the effort to write the patch) then why not?

(And yeah, yeah, someone might find a .gif of a parrot useful too. Go do some work.)

#8 lore raw
In reply to Pavel Machek
Quoted 17 lines · On Wed, 26 Nov 2008 17:15:21 +0100 Pavel Machek wrote:

Document how to deal with bad memory reported with memtest.

Documentation/bad_memory.txt new file +7 -0
new file mode 100644
index 0000000..df84162
@@ -0,0 +1,45 @@
+March 2008
+Jan-Simon Moeller, [email protected]
+
+
+How to deal with bad memory e.g. reported by memtest86+ ?
+#########################################################
+

Documentation/bad_memory.txt line 17

Quoted 10 lines
Documentation/bad_memory.txt new file +10 -0
+There are three possibilities I know of:
+
+1) Reinsert/swap the memory modules
+
+2) Buy new modules (best!) or try to exchange the memory
+   if you have spare-parts
+
+3) Use BadRAM or memmap
+
+This Howto is about number 3) .

No space between 3) and '.'.

Documentation/bad_memory.txt line 30

Quoted 13 lines
Documentation/bad_memory.txt new file +13 -0
+
+
+BadRAM
+######
+BadRAM is the actively developed and available as kernel-patch
+here:  http://rick.vanrein.org/linux/badram/
+
+For more details see the BadRAM documentation.
+
+memmap
+######
+
+memmap is already in the kernel and usable as kernel-parameter at

a kernel parameter at

Documentation/bad_memory.txt line 31

Quoted 1 line
Documentation/bad_memory.txt new file +1 -0
+boot-time.  Its syntax is slightly strange and you may need to

boot time.

Documentation/bad_memory.txt line 32

Quoted 1 line
Documentation/bad_memory.txt new file +1 -0
+calculate the values by yourself!

s/!/./

Documentation/bad_memory.txt line 37

Quoted 5 lines
Documentation/bad_memory.txt new file +5 -0
+
+Syntax to exclude a memory area (see kernel-parameters.txt for details):
+memmap=<size>$<address>
+
+Example: memtest86+ reported here errors at address 0x18691458, 0x18698424 and

s/here //

Documentation/bad_memory.txt line 39

Quoted 2 lines
Documentation/bad_memory.txt new file +2 -0
+         some others. All had 0x1869xxxx in common, so I chose a pattern of
+         0x18690000,0xffff0000.

What is the 0xffff0000 for? Needs explanation.

Documentation/bad_memory.txt line 45

Quoted 6 lines
Documentation/bad_memory.txt new file +6 -0
+
+With the numbers of the example above:
+memmap=64K$0x18690000
+ or
+memmap=0x10000$0x18690000
+

Please lose the last empty line.

and thanks for the patch/new file.

---
~Randy

#9 lore raw

I cleaned the document up according to Randy (thanks!). I don't actually know enough about DRAM error characcteristics, I guess'round the size of bad region up to nearest 2^n makes sense.

Documentation/bad_memory.txt +9 -8
index df84162..a2a8703 100644
@@ -14,12 +14,12 @@ There are three possibilities I know of:
 
 3) Use BadRAM or memmap
 
-This Howto is about number 3) .
+This Howto is about number 3).
 
 
 BadRAM
 ######
-BadRAM is the actively developed and available as kernel-patch
+BadRAM is the actively developed and available as a kernel patch
 here:  http://rick.vanrein.org/linux/badram/
 
 For more details see the BadRAM documentation.
@@ -27,19 +27,20 @@ For more details see the BadRAM documentation.
 memmap
 ######
 
-memmap is already in the kernel and usable as kernel-parameter at
-boot-time.  Its syntax is slightly strange and you may need to
-calculate the values by yourself!
+memmap is already in the kernel and usable as a kernel parameter at
+boot time.  Its syntax is slightly strange and you may need to
+calculate the values by yourself.
 
 Syntax to exclude a memory area (see kernel-parameters.txt for details):
 memmap=<size>$<address>
 
-Example: memtest86+ reported here errors at address 0x18691458, 0x18698424 and
+Example: memtest86+ reported errors at address 0x18691458, 0x18698424 and
          some others. All had 0x1869xxxx in common, so I chose a pattern of
-         0x18690000,0xffff0000.
+         0x18690000 and size of 0x10000. (Size needs to cover at least all
+	 known bad places, and rounding to nearest power of 2 makes sense
+	 'just to be safe').
 
 With the numbers of the example above:
 memmap=64K$0x18690000
  or
 memmap=0x10000$0x18690000
-
Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
#10 lore raw
Quoted 7 lines · On Tuesday 09 December 2008 06:31:52 Pavel Machek wrote:

I cleaned the document up according to Randy (thanks!). I don't actually know enough about DRAM error characcteristics, I guess'round the size of bad region up to nearest 2^n makes sense.

Documentation/bad_memory.txt +0 -0

...

Documentation/bad_memory.txt There are three possibilities I know of: · line 23

Quoted 8 lines
Documentation/bad_memory.txt +2 -1
+This Howto is about number 3).
 
 
 BadRAM
 ######
-BadRAM is the actively developed and available as kernel-patch
+BadRAM is the actively developed and available as a kernel patch
 here:  http://rick.vanrein.org/linux/badram/

Ok, once again: the point of this patch is to document an out of tree patch.

The out of tree patch is here:
http://rick.vanrein.org/linux/badram/software/BadRAM-2.6.27.1.patch

It has its own Documentation/badram.txt file and it patches Documentation/memory.txt, as acknowledged here:

Documentation/bad_memory.txt For more details see the BadRAM documentation. · line 28

Quoted 4 lines

For more details see the BadRAM documentation.

Documentation/bad_memory.txt +0 -0
@@ -27,19 +27,20 @@ For more details see the BadRAM documentation.
 memmap
 ######

Now what I don't understand is, why add something to the tree formalizing the out-of-tree status of this other patch? Why not just merge it? If it's interesting enough to have documentation about the patch in the tree, why is the patch itself not interesting enough to merge? It's clearly got an active maintainer, and has for years. (Is there something specific about it that needs to be cleaned up?)

Adding this extra documentation to the badram patch sounds great. Merging the badram patch into the linux kernel sounds useful; obviously _this_ patch is inherently an expression of interest in it. Adding documentation about the badram patch to the linux kernel tree but _not_ adding the badram patch itself seems kind of crazy.

Would someone please explain the reasoning here? I don't understand it.

Rob

#11 lore raw
Quoted 12 lines · Rob Landley
Quoted 7 lines · On Tuesday 09 December 2008 06:31:52 Pavel Machek wrote:

I cleaned the document up according to Randy (thanks!). I don't actually know enough about DRAM error characcteristics, I guess'round the size of bad region up to nearest 2^n makes sense.

Documentation/bad_memory.txt +0 -0

...

Quoted 3 lines

+This Howto is about number 3).

Quoted 7 lines

Documentation/bad_memory.txt There are three possibilities I know of: · line 23

Quoted 5 lines

BadRAM
######

Documentation/bad_memory.txt +1 -1
-BadRAM is the actively developed and available as kernel-patch
+BadRAM is the actively developed and available as a kernel patch
 here:  http://rick.vanrein.org/linux/badram/

Ok, once again: the point of this patch is to document an out of tree patch.

No; the point of this piece of documentation is to tell people how to work _without_ that patch. Because it is simple enough.

Quoted 6 lines

The out of tree patch is here:
http://rick.vanrein.org/linux/badram/software/BadRAM-2.6.27.1.patch

It has its own Documentation/badram.txt file and it patches Documentation/memory.txt, as acknowledged here:

Quoted 8 lines

Documentation/bad_memory.txt For more details see the BadRAM documentation. · line 28

Quoted 4 lines

For more details see the BadRAM documentation.

Documentation/bad_memory.txt +0 -0
@@ -27,19 +27,20 @@ For more details see the BadRAM documentation.
 memmap
 ######

Now what I don't understand is, why add something to the tree formalizing the out-of-tree status of this other patch? Why not just merge it? If it's

Take a look at that patch. It is seriously overengineered. This should not need a config option, should not introduce new page flag, etc.

We already have perfectly working interface for excluding specific addresses; maybe we need better documentation, and maybe kernel commandline interface should be changed to be more user friendly, but we certainly don't want to take the badram patch.

This excerpt should be enough:

diff -pruN linux-2.6.27/include/linux/page-flags.h linux-2.6.27-new/include/linux/page-flags.h
--- linux-2.6.27/include/linux/page-flags.h 2008-10-10
03:43:53.000000000 +0530
+++ linux-2.6.27-new/include/linux/page-flags.h 2008-10-15
10:04:48.000000000 +0530
@@ -93,6 +93,9 @@ enum pageflags {
PG_mappedtodisk, /* Has blocks allocated on-disk */
PG_reclaim, /* To be reclaimed asap */
PG_buddy, /* Page is free, on buddy lists */
+#ifdef CONFIG_BADRAM
+ PG_badram, /* BadRam page */
+#endif
#ifdef CONFIG_IA64_UNCACHED_ALLOCATOR
PG_uncached, /* Page has been mapped as uncached */
#

Pavel

Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html