rpms
/
kernel
2
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

edac: Fix memory filling order 

After testing it on a machine with in asynchronous mode, it was
noticed that the memories are not filled in the proper order.
Fix it.

Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
This commit is contained in:
Mauro Carvalho Chehab 2012-10-24 10:32:10 -02:00
parent 8e8972af61
commit c15eb72f4b
2 changed files with 101 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

98
i82975x-edac-fix.patch
View File

@ -853,3 +853,101 @@ index f998d2c..082e91e 100644
fail0:
return rc;
}
commit ad28406017c4f225f10cb78931214e48fde866ae
Author: Mauro Carvalho Chehab <mchehab@redhat.com>
Date: Wed Oct 24 10:30:01 2012 -0200
edac: Fix the dimm filling for csrows-based layouts
The driver is currently filling data in a wrong way, on drivers
for csrows-based memory controller, when the first layer is a
csrow.
This is not easily to notice, as, in general, memories are
filed in dual, interleaved, symetric mode, as very few memory
controllers support asymetric modes.
While digging into a bug for i82795_edac driver, the asymetric
mode there is now working, allowing us to fill the machine with
4x1GB ranks at channel 0, and 2x512GB at channel 1:
Channel 0 ranks:
EDAC DEBUG: i82975x_init_csrows: DIMM A0: from page 0x00000000 to 0x0003ffff (size: 0x00040000 pages)
EDAC DEBUG: i82975x_init_csrows: DIMM A1: from page 0x00040000 to 0x0007ffff (size: 0x00040000 pages)
EDAC DEBUG: i82975x_init_csrows: DIMM A2: from page 0x00080000 to 0x000bffff (size: 0x00040000 pages)
EDAC DEBUG: i82975x_init_csrows: DIMM A3: from page 0x000c0000 to 0x000fffff (size: 0x00040000 pages)
Channel 1 ranks:
EDAC DEBUG: i82975x_init_csrows: DIMM B0: from page 0x00100000 to 0x0011ffff (size: 0x00020000 pages)
EDAC DEBUG: i82975x_init_csrows: DIMM B1: from page 0x00120000 to 0x0013ffff (size: 0x00020000 pages)
Instead of properly showing the memories as such, before this patch, it
shows the memory layout as:
+-----------------------------------+
| mc0 |
| csrow0 | csrow1 | csrow2 |
----------+-----------------------------------+
channel1: | 1024 MB | 1024 MB | 512 MB |
channel0: | 1024 MB | 1024 MB | 512 MB |
----------+-----------------------------------+
as if both channels were symetric, grouping the DIMMs on a wrong
layout.
After this patch, the memory is correctly represented.
So, for csrows at layers[0], it shows:
+-----------------------------------------------+
| mc0 |
| csrow0 | csrow1 | csrow2 | csrow3 |
----------+-----------------------------------------------+
channel1: | 512 MB | 512 MB | 0 MB | 0 MB |
channel0: | 1024 MB | 1024 MB | 1024 MB | 1024 MB |
----------+-----------------------------------------------+
For csrows at layers[1], it shows:
+-----------------------+
| mc0 |
| channel0 | channel1 |
--------+-----------------------+
csrow3: | 1024 MB | 0 MB |
csrow2: | 1024 MB | 0 MB |
--------+-----------------------+
csrow1: | 1024 MB | 512 MB |
csrow0: | 1024 MB | 512 MB |
--------+-----------------------+
So, no matter of what comes first, the information between
channel and csrow will be properly represented.
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
diff --git a/drivers/edac/edac_mc.c b/drivers/edac/edac_mc.c
index d5dc9da..81eb9fd 100644
--- a/drivers/edac/edac_mc.c
+++ b/drivers/edac/edac_mc.c
@@ -416,10 +416,18 @@ struct mem_ctl_info *edac_mc_alloc(unsigned mc_num,
dimm->cschannel = chn;
/* Increment csrow location */
- row++;
- if (row == tot_csrows) {
- row = 0;
+ if (layers[0].is_virt_csrow) {
chn++;
+ if (chn == tot_channels) {
+ chn = 0;
+ row++;
+ }
+ } else {
+ row++;
+ if (row == tot_csrows) {
+ row = 0;
+ chn++;
+ }
}
/* Increment dimm location */

3
kernel.spec
View File

@ -2360,6 +2360,9 @@ fi
# '-' | |
# '-'
%changelog
* Tue Oct 30 2012 Mauro Carvalho Chehab <mchehab@redhat.com>
- Fix EDAC memory filling for csrow-based memory controllers
* Tue Oct 30 2012 Josh Boyer <jwboyer@redhat.com>
- Move power-x86-destdir.patch to apply on vanilla kernels (thanks knurd)