* Memory controller not showing half of the memory?
@ 2025-03-13 17:21 reox
2025-03-13 17:31 ` Borislav Petkov
0 siblings, 1 reply; 13+ messages in thread
From: reox @ 2025-03-13 17:21 UTC (permalink / raw)
To: linux-edac
Hello,
I hope it is okay to ask such a question on this mailinglist, however
there is surprisingly little information on this topic (or I'm not deep
enough in the matter to understand everything correctly...).
I have a server with a ASRockRack B665D4U-1L board. I'm running Debian
with bookworm-backports kernel 6.12.9-1~bpo12+1.
In the syslog I see the following output of EDAC:
# dmesg | grep -i EDAC
[ 1.340448] EDAC MC: Ver: 3.0.0
[ 4.705255] EDAC MC0: Giving out device to module amd64_edac
controller F19h_M70h: DEV 0000:00:18.3 (INTERRUPT)
[ 4.705262] EDAC amd64: F19h_M70h detected (node 0).
[ 4.705266] EDAC MC: UMC0 chip selects:
[ 4.705267] EDAC amd64: MC: 0: 0MB 1: 0MB
[ 4.705270] EDAC amd64: MC: 2: 8192MB 3: 8192MB
[ 4.705273] EDAC MC: UMC1 chip selects:
[ 4.705274] EDAC amd64: MC: 0: 0MB 1: 0MB
[ 4.705277] EDAC amd64: MC: 2: 8192MB 3: 8192MB
However, compare this to the real amount of memory (2×32GB):
# lshw -c memory # shorted
*-memory
size: 64GiB
capabilities: ecc
configuration: errordetection=multi-bit-ecc
*-bank:0
description: [empty]
*-bank:1
description: DIMM Synchronous Unbuffered (Unregistered) 4800
MHz (0.2 ns)
product: HMCG88MEBEA081N
vendor: SK Hynix
size: 32GiB
*-bank:2
description: [empty]
*-bank:3
description: DIMM Synchronous Unbuffered (Unregistered) 4800
MHz (0.2 ns)
product: HMCG88MEBEA081N
vendor: SK Hynix
size: 32GiB
I'm a bit confused here... Does EDAC simply miss half of the memory? Is
this output correct? Or the memory controller not fully implemented /
bugged in that kernel version (with the bookworm kernel it does not load
EDAC at all)? Unfortunately I cannot easily run a dev version of the
kernel on that machine - otherwise I would have probably already done that.
Thanks in advance!
Best,
Sebastian
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: Memory controller not showing half of the memory?
2025-03-13 17:21 Memory controller not showing half of the memory? reox
@ 2025-03-13 17:31 ` Borislav Petkov
2025-03-13 17:48 ` reox
0 siblings, 1 reply; 13+ messages in thread
From: Borislav Petkov @ 2025-03-13 17:31 UTC (permalink / raw)
To: reox; +Cc: linux-edac, Yazen Ghannam
On Thu, Mar 13, 2025 at 06:21:28PM +0100, reox wrote:
> Hello,
>
> I hope it is okay to ask such a question on this mailinglist, however there
> is surprisingly little information on this topic (or I'm not deep enough in
> the matter to understand everything correctly...).
>
> I have a server with a ASRockRack B665D4U-1L board. I'm running Debian with
> bookworm-backports kernel 6.12.9-1~bpo12+1.
> In the syslog I see the following output of EDAC:
> # dmesg | grep -i EDAC
> [ 1.340448] EDAC MC: Ver: 3.0.0
> [ 4.705255] EDAC MC0: Giving out device to module amd64_edac controller
> F19h_M70h: DEV 0000:00:18.3 (INTERRUPT)
> [ 4.705262] EDAC amd64: F19h_M70h detected (node 0).
> [ 4.705266] EDAC MC: UMC0 chip selects:
> [ 4.705267] EDAC amd64: MC: 0: 0MB 1: 0MB
> [ 4.705270] EDAC amd64: MC: 2: 8192MB 3: 8192MB
> [ 4.705273] EDAC MC: UMC1 chip selects:
> [ 4.705274] EDAC amd64: MC: 0: 0MB 1: 0MB
> [ 4.705277] EDAC amd64: MC: 2: 8192MB 3: 8192MB
>
> However, compare this to the real amount of memory (2×32GB):
> # lshw -c memory # shorted
> *-memory
> size: 64GiB
> capabilities: ecc
> configuration: errordetection=multi-bit-ecc
> *-bank:0
> description: [empty]
> *-bank:1
> description: DIMM Synchronous Unbuffered (Unregistered) 4800 MHz
> (0.2 ns)
> product: HMCG88MEBEA081N
> vendor: SK Hynix
> size: 32GiB
> *-bank:2
> description: [empty]
> *-bank:3
> description: DIMM Synchronous Unbuffered (Unregistered) 4800 MHz
> (0.2 ns)
> product: HMCG88MEBEA081N
> vendor: SK Hynix
> size: 32GiB
>
> I'm a bit confused here... Does EDAC simply miss half of the memory? Is this
> output correct? Or the memory controller not fully implemented / bugged in
> that kernel version (with the bookworm kernel it does not load EDAC at all)?
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
What does that mean, exactly?
Are you force-loading the module?
You could send a full dmesg...
> Unfortunately I cannot easily run a dev version of the kernel on that
> machine - otherwise I would have probably already done that.
>
> Thanks in advance!
> Best,
> Sebastian
>
--
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquette
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: Memory controller not showing half of the memory?
2025-03-13 17:31 ` Borislav Petkov
@ 2025-03-13 17:48 ` reox
2025-03-17 12:07 ` Yazen Ghannam
0 siblings, 1 reply; 13+ messages in thread
From: reox @ 2025-03-13 17:48 UTC (permalink / raw)
To: Borislav Petkov; +Cc: linux-edac, Yazen Ghannam
Am 13.03.2025 um 18:31 schrieb Borislav Petkov:
> On Thu, Mar 13, 2025 at 06:21:28PM +0100, reox wrote:
>> Hello,
>>
>> I hope it is okay to ask such a question on this mailinglist, however there
>> is surprisingly little information on this topic (or I'm not deep enough in
>> the matter to understand everything correctly...).
>>
>> I have a server with a ASRockRack B665D4U-1L board. I'm running Debian with
>> bookworm-backports kernel 6.12.9-1~bpo12+1.
>> In the syslog I see the following output of EDAC:
>> # dmesg | grep -i EDAC
>> [ 1.340448] EDAC MC: Ver: 3.0.0
>> [ 4.705255] EDAC MC0: Giving out device to module amd64_edac controller
>> F19h_M70h: DEV 0000:00:18.3 (INTERRUPT)
>> [ 4.705262] EDAC amd64: F19h_M70h detected (node 0).
>> [ 4.705266] EDAC MC: UMC0 chip selects:
>> [ 4.705267] EDAC amd64: MC: 0: 0MB 1: 0MB
>> [ 4.705270] EDAC amd64: MC: 2: 8192MB 3: 8192MB
>> [ 4.705273] EDAC MC: UMC1 chip selects:
>> [ 4.705274] EDAC amd64: MC: 0: 0MB 1: 0MB
>> [ 4.705277] EDAC amd64: MC: 2: 8192MB 3: 8192MB
>>
>> However, compare this to the real amount of memory (2×32GB):
>> # lshw -c memory # shorted
>> *-memory
>> size: 64GiB
>> capabilities: ecc
>> configuration: errordetection=multi-bit-ecc
>> *-bank:0
>> description: [empty]
>> *-bank:1
>> description: DIMM Synchronous Unbuffered (Unregistered) 4800 MHz
>> (0.2 ns)
>> product: HMCG88MEBEA081N
>> vendor: SK Hynix
>> size: 32GiB
>> *-bank:2
>> description: [empty]
>> *-bank:3
>> description: DIMM Synchronous Unbuffered (Unregistered) 4800 MHz
>> (0.2 ns)
>> product: HMCG88MEBEA081N
>> vendor: SK Hynix
>> size: 32GiB
>>
>> I'm a bit confused here... Does EDAC simply miss half of the memory? Is this
>> output correct? Or the memory controller not fully implemented / bugged in
>> that kernel version (with the bookworm kernel it does not load EDAC at all)?
> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>
> What does that mean, exactly?
>
> Are you force-loading the module?
Sorry, that was formulated a bit sloppy... I meant that with the current
bookworm kernel (6.1.128) there was no memory controller detected. The
module was loaded fine but reported nothing. As far as I can tell,
support was added only with 6.5 with commit 6c79e42 (would the expected
outcome in my case also be 4×16GB?).
by the way, if it helps, the CPU is a "AMD Ryzen 7 PRO 8700GE w/ Radeon
780M Graphics".
>
> You could send a full dmesg...
from 6.1.128 or 6.12.9?
>
>> Unfortunately I cannot easily run a dev version of the kernel on that
>> machine - otherwise I would have probably already done that.
>>
>> Thanks in advance!
>> Best,
>> Sebastian
>>
>
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: Memory controller not showing half of the memory?
2025-03-13 17:48 ` reox
@ 2025-03-17 12:07 ` Yazen Ghannam
2025-03-17 12:37 ` reox
2025-03-20 9:38 ` Žilvinas Žaltiena
0 siblings, 2 replies; 13+ messages in thread
From: Yazen Ghannam @ 2025-03-17 12:07 UTC (permalink / raw)
To: reox; +Cc: Borislav Petkov, linux-edac
On Thu, Mar 13, 2025 at 06:48:42PM +0100, reox wrote:
>
>
> Am 13.03.2025 um 18:31 schrieb Borislav Petkov:
> > On Thu, Mar 13, 2025 at 06:21:28PM +0100, reox wrote:
> > > Hello,
> > >
> > > I hope it is okay to ask such a question on this mailinglist, however there
> > > is surprisingly little information on this topic (or I'm not deep enough in
> > > the matter to understand everything correctly...).
> > >
> > > I have a server with a ASRockRack B665D4U-1L board. I'm running Debian with
> > > bookworm-backports kernel 6.12.9-1~bpo12+1.
> > > In the syslog I see the following output of EDAC:
> > > # dmesg | grep -i EDAC
> > > [ 1.340448] EDAC MC: Ver: 3.0.0
> > > [ 4.705255] EDAC MC0: Giving out device to module amd64_edac controller
> > > F19h_M70h: DEV 0000:00:18.3 (INTERRUPT)
> > > [ 4.705262] EDAC amd64: F19h_M70h detected (node 0).
> > > [ 4.705266] EDAC MC: UMC0 chip selects:
> > > [ 4.705267] EDAC amd64: MC: 0: 0MB 1: 0MB
> > > [ 4.705270] EDAC amd64: MC: 2: 8192MB 3: 8192MB
> > > [ 4.705273] EDAC MC: UMC1 chip selects:
> > > [ 4.705274] EDAC amd64: MC: 0: 0MB 1: 0MB
> > > [ 4.705277] EDAC amd64: MC: 2: 8192MB 3: 8192MB
> > >
> > > However, compare this to the real amount of memory (2×32GB):
> > > # lshw -c memory # shorted
> > > *-memory
> > > size: 64GiB
> > > capabilities: ecc
> > > configuration: errordetection=multi-bit-ecc
> > > *-bank:0
> > > description: [empty]
> > > *-bank:1
> > > description: DIMM Synchronous Unbuffered (Unregistered) 4800 MHz
> > > (0.2 ns)
> > > product: HMCG88MEBEA081N
> > > vendor: SK Hynix
> > > size: 32GiB
> > > *-bank:2
> > > description: [empty]
> > > *-bank:3
> > > description: DIMM Synchronous Unbuffered (Unregistered) 4800 MHz
> > > (0.2 ns)
> > > product: HMCG88MEBEA081N
> > > vendor: SK Hynix
> > > size: 32GiB
> > >
> > > I'm a bit confused here... Does EDAC simply miss half of the memory? Is this
> > > output correct? Or the memory controller not fully implemented / bugged in
> > > that kernel version (with the bookworm kernel it does not load EDAC at all)?
> > ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> >
> > What does that mean, exactly?
> >
> > Are you force-loading the module?
>
> Sorry, that was formulated a bit sloppy... I meant that with the current
> bookworm kernel (6.1.128) there was no memory controller detected. The
> module was loaded fine but reported nothing. As far as I can tell, support
> was added only with 6.5 with commit 6c79e42 (would the expected outcome in
> my case also be 4×16GB?).
>
> by the way, if it helps, the CPU is a "AMD Ryzen 7 PRO 8700GE w/ Radeon 780M
> Graphics".
>
> >
> > You could send a full dmesg...
>
> from 6.1.128 or 6.12.9?
>
From the latest, please.
It could be that the memory layout/assignment is slightly different on
this model group compared to others. I'll see what I can find out.
Thanks for reporting this!
Thanks,
Yazen
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: Memory controller not showing half of the memory?
2025-03-17 12:07 ` Yazen Ghannam
@ 2025-03-17 12:37 ` reox
2025-03-20 9:38 ` Žilvinas Žaltiena
1 sibling, 0 replies; 13+ messages in thread
From: reox @ 2025-03-17 12:37 UTC (permalink / raw)
To: Yazen Ghannam; +Cc: Borislav Petkov, linux-edac
[-- Attachment #1: Type: text/plain, Size: 3367 bytes --]
Hi,
Am 17.03.2025 um 13:07 schrieb Yazen Ghannam:
> On Thu, Mar 13, 2025 at 06:48:42PM +0100, reox wrote:
>>
>>
>> Am 13.03.2025 um 18:31 schrieb Borislav Petkov:
>>> On Thu, Mar 13, 2025 at 06:21:28PM +0100, reox wrote:
>>>> Hello,
>>>>
>>>> I hope it is okay to ask such a question on this mailinglist, however there
>>>> is surprisingly little information on this topic (or I'm not deep enough in
>>>> the matter to understand everything correctly...).
>>>>
>>>> I have a server with a ASRockRack B665D4U-1L board. I'm running Debian with
>>>> bookworm-backports kernel 6.12.9-1~bpo12+1.
>>>> In the syslog I see the following output of EDAC:
>>>> # dmesg | grep -i EDAC
>>>> [ 1.340448] EDAC MC: Ver: 3.0.0
>>>> [ 4.705255] EDAC MC0: Giving out device to module amd64_edac controller
>>>> F19h_M70h: DEV 0000:00:18.3 (INTERRUPT)
>>>> [ 4.705262] EDAC amd64: F19h_M70h detected (node 0).
>>>> [ 4.705266] EDAC MC: UMC0 chip selects:
>>>> [ 4.705267] EDAC amd64: MC: 0: 0MB 1: 0MB
>>>> [ 4.705270] EDAC amd64: MC: 2: 8192MB 3: 8192MB
>>>> [ 4.705273] EDAC MC: UMC1 chip selects:
>>>> [ 4.705274] EDAC amd64: MC: 0: 0MB 1: 0MB
>>>> [ 4.705277] EDAC amd64: MC: 2: 8192MB 3: 8192MB
>>>>
>>>> However, compare this to the real amount of memory (2×32GB):
>>>> # lshw -c memory # shorted
>>>> *-memory
>>>> size: 64GiB
>>>> capabilities: ecc
>>>> configuration: errordetection=multi-bit-ecc
>>>> *-bank:0
>>>> description: [empty]
>>>> *-bank:1
>>>> description: DIMM Synchronous Unbuffered (Unregistered) 4800 MHz
>>>> (0.2 ns)
>>>> product: HMCG88MEBEA081N
>>>> vendor: SK Hynix
>>>> size: 32GiB
>>>> *-bank:2
>>>> description: [empty]
>>>> *-bank:3
>>>> description: DIMM Synchronous Unbuffered (Unregistered) 4800 MHz
>>>> (0.2 ns)
>>>> product: HMCG88MEBEA081N
>>>> vendor: SK Hynix
>>>> size: 32GiB
>>>>
>>>> I'm a bit confused here... Does EDAC simply miss half of the memory? Is this
>>>> output correct? Or the memory controller not fully implemented / bugged in
>>>> that kernel version (with the bookworm kernel it does not load EDAC at all)?
>>> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>>>
>>> What does that mean, exactly?
>>>
>>> Are you force-loading the module?
>>
>> Sorry, that was formulated a bit sloppy... I meant that with the current
>> bookworm kernel (6.1.128) there was no memory controller detected. The
>> module was loaded fine but reported nothing. As far as I can tell, support
>> was added only with 6.5 with commit 6c79e42 (would the expected outcome in
>> my case also be 4×16GB?).
>>
>> by the way, if it helps, the CPU is a "AMD Ryzen 7 PRO 8700GE w/ Radeon 780M
>> Graphics".
>>
>>>
>>> You could send a full dmesg...
>>
>> from 6.1.128 or 6.12.9?
>>
>
> From the latest, please.
>
> It could be that the memory layout/assignment is slightly different on
> this model group compared to others. I'll see what I can find out.
Sure, it is attached. This is with 6.12.12 now.
It is a bit hard to install arbitrary kernel versions on that machine,
as it is in production, however, I can try my best to test other kernels
if required.
Best,
Sebastian
>
> Thanks for reporting this!
>
> Thanks,
> Yazen
>
[-- Attachment #2: dmesg.log.zst --]
[-- Type: application/x-compressed, Size: 21949 bytes --]
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: Memory controller not showing half of the memory?
2025-03-17 12:07 ` Yazen Ghannam
2025-03-17 12:37 ` reox
@ 2025-03-20 9:38 ` Žilvinas Žaltiena
2025-03-28 15:24 ` Borislav Petkov
1 sibling, 1 reply; 13+ messages in thread
From: Žilvinas Žaltiena @ 2025-03-20 9:38 UTC (permalink / raw)
To: yazen.ghannam; +Cc: bp, linux-edac, mailinglist
Hello,
I am also having this "half" memory problem - EDAC is showing 48GB
instead of 96GB, however otherwise whole 96GB are usable by OS.
My system has:
1) AMD Ryzen 7950X3D
2) Asus B650 Pro Creator motherboard (AGESA is PI 1.2.0.3a Patch A)
3) 2x48GB ECC UDIMMs (KSM56E46BD8KM-48HM
https://www.kingston.com/datasheets/KSM56E46BD8KM-48HM.pdf)
Kernel 6.13.6 is showing:
[ 0.325894] [ T1] EDAC MC: Ver: 3.0.0
[ 9.521853] [ T1336] EDAC MC0: Giving out device to module
amd64_edac controller F19h_M60h: DEV 0000:00:18.3 (INTERRUPT)
[ 9.521856] [ T1336] EDAC amd64: F19h_M60h detected (node 0).
[ 9.521857] [ T1336] EDAC MC: UMC0 chip selects:
[ 9.521858] [ T1336] EDAC amd64: MC: 0: 0MB 1: 0MB
[ 9.521859] [ T1336] EDAC amd64: MC: 2: 16384MB 3: 8192MB
[ 9.521860] [ T1336] EDAC MC: UMC1 chip selects:
[ 9.521861] [ T1336] EDAC amd64: MC: 0: 0MB 1: 0MB
[ 9.521862] [ T1336] EDAC amd64: MC: 2: 16384MB 3: 8192MB
or in summary:
UMC0: MC0=0GB, MC1=0GB, MC2=16GB, MC3=8GB
UMC1: MC0=0GB, MC1=0GB, MC2=16GB, MC3=8GB
I also has access to different system with AMD Ryzen 7950X and 2x32GB
ECC UDIMMs, and it does NOT have this half ram problem, i.e. EDAC shows
in summary:
UMC0: MC0=0GB, MC1=0GB, MC2=16GB, MC3=16GB
UMC1: MC0=0GB, MC1=0GB, MC2=16GB, MC3=16GB
RAM is Kingston KSM52E42BD8KM-32HA
(https://www.kingston.com/datasheets/KSM52E42BD8KM-32HA.pdf) in it.
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: Memory controller not showing half of the memory?
2025-03-20 9:38 ` Žilvinas Žaltiena
@ 2025-03-28 15:24 ` Borislav Petkov
2025-03-30 8:25 ` Žilvinas
2025-04-01 11:44 ` mailinglist
0 siblings, 2 replies; 13+ messages in thread
From: Borislav Petkov @ 2025-03-28 15:24 UTC (permalink / raw)
To: Žilvinas Žaltiena, mailinglist
Cc: yazen.ghannam, linux-edac, Avadhut Naik
On Thu, Mar 20, 2025 at 11:38:58AM +0200, Žilvinas Žaltiena wrote:
> Hello,
>
> I am also having this "half" memory problem - EDAC is showing 48GB instead
> of 96GB, however otherwise whole 96GB are usable by OS.
Both of y'all folks encountering this issue, care to run the below patch?
https://lore.kernel.org/all/20250327210718.1640762-1-avadhut.naik@amd.com/
Thx.
--
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquette
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: Memory controller not showing half of the memory?
2025-03-28 15:24 ` Borislav Petkov
@ 2025-03-30 8:25 ` Žilvinas
2025-03-30 10:39 ` Borislav Petkov
2025-04-01 11:44 ` mailinglist
1 sibling, 1 reply; 13+ messages in thread
From: Žilvinas @ 2025-03-30 8:25 UTC (permalink / raw)
To: Borislav Petkov, mailinglist; +Cc: yazen.ghannam, linux-edac, Avadhut Naik
On 2025-03-28 17:24, Borislav Petkov wrote:
> Both of y'all folks encountering this issue, care to run the below patch?
>
> https://lore.kernel.org/all/20250327210718.1640762-1-avadhut.naik@amd.com/
>
> Thx.
>
I have applied that patch on top of 6.13.8 and now EDAC shows:
UMC0: MC0=0GB, MC1=0GB, MC2=24GB, MC3=24GB
UMC1: MC0=0GB, MC1=0GB, MC2=24GB, MC3=24GB
it looks correct now.
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: Memory controller not showing half of the memory?
2025-03-30 8:25 ` Žilvinas
@ 2025-03-30 10:39 ` Borislav Petkov
2025-03-30 11:00 ` Žilvinas Žaltiena
0 siblings, 1 reply; 13+ messages in thread
From: Borislav Petkov @ 2025-03-30 10:39 UTC (permalink / raw)
To: Žilvinas; +Cc: mailinglist, yazen.ghannam, linux-edac, Avadhut Naik
On Sun, Mar 30, 2025 at 11:25:17AM +0300, Žilvinas wrote:
> I have applied that patch on top of 6.13.8 and now EDAC shows:
>
> UMC0: MC0=0GB, MC1=0GB, MC2=24GB, MC3=24GB
> UMC1: MC0=0GB, MC1=0GB, MC2=24GB, MC3=24GB
>
> it looks correct now.
Thanks for testing.
Want me to add your tags to the patch:
Reported-by: Žilvinas Žaltiena <zilvinas@natrix.lt>
Tested-by: Žilvinas Žaltiena <zilvinas@natrix.lt>
?
--
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquette
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: Memory controller not showing half of the memory?
2025-03-30 10:39 ` Borislav Petkov
@ 2025-03-30 11:00 ` Žilvinas Žaltiena
0 siblings, 0 replies; 13+ messages in thread
From: Žilvinas Žaltiena @ 2025-03-30 11:00 UTC (permalink / raw)
To: Borislav Petkov; +Cc: mailinglist, yazen.ghannam, linux-edac, Avadhut Naik
On 2025-03-30 13:39, Borislav Petkov wrote:
> Want me to add your tags to the patch:
>
> Reported-by: Žilvinas Žaltiena <zilvinas@natrix.lt>
> Tested-by: Žilvinas Žaltiena <zilvinas@natrix.lt>
>
> ?
>
OK.
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: Memory controller not showing half of the memory?
2025-03-28 15:24 ` Borislav Petkov
2025-03-30 8:25 ` Žilvinas
@ 2025-04-01 11:44 ` mailinglist
2025-04-01 18:00 ` Naik, Avadhut
1 sibling, 1 reply; 13+ messages in thread
From: mailinglist @ 2025-04-01 11:44 UTC (permalink / raw)
To: Borislav Petkov, Žilvinas Žaltiena
Cc: yazen.ghannam, linux-edac, Avadhut Naik
Am 28.03.2025 um 16:24 schrieb Borislav Petkov:
> On Thu, Mar 20, 2025 at 11:38:58AM +0200, Žilvinas Žaltiena wrote:
>> Hello,
>>
>> I am also having this "half" memory problem - EDAC is showing 48GB instead
>> of 96GB, however otherwise whole 96GB are usable by OS.
>
> Both of y'all folks encountering this issue, care to run the below patch?
>
> https://lore.kernel.org/all/20250327210718.1640762-1-avadhut.naik@amd.com/
>
> Thx.
>
I tried to apply it on the current Bookworm Backports kernel (6.12.12) -
which seemed to work but I still get the same output, i.e., 32GB out of
64GB of RAM.
I may not have built the kernel correct though... I did not had an
environment to do that and just set one up. Or do I need a newer kernel
version to apply the patch to? Sorry for the noob questions...
Best,
Sebastian
^ permalink raw reply [flat|nested] 13+ messages in thread
* Memory controller not showing half of the memory?
2025-04-01 11:44 ` mailinglist
@ 2025-04-01 18:00 ` Naik, Avadhut
2025-07-24 17:47 ` mailinglist
0 siblings, 1 reply; 13+ messages in thread
From: Naik, Avadhut @ 2025-04-01 18:00 UTC (permalink / raw)
To: mailinglist, Žilvinas Žaltiena
Cc: yazen.ghannam, linux-edac, Borislav Petkov
On 4/1/2025 06:44, mailinglist@reox.at wrote:
>
>
> Am 28.03.2025 um 16:24 schrieb Borislav Petkov:
>> On Thu, Mar 20, 2025 at 11:38:58AM +0200, Žilvinas Žaltiena wrote:
>>> Hello,
>>>
>>> I am also having this "half" memory problem - EDAC is showing 48GB instead
>>> of 96GB, however otherwise whole 96GB are usable by OS.
>>
>> Both of y'all folks encountering this issue, care to run the below patch?
>>
>> https://lore.kernel.org/all/20250327210718.1640762-1-avadhut.naik@amd.com/
>>
>> Thx.
>>
>
> I tried to apply it on the current Bookworm Backports kernel (6.12.12) - which seemed to work but I still get the same output, i.e., 32GB out of 64GB of RAM.
>
Think we are mixing up issues here. So, to clarify, there are two issues currently being worked on in the
amd64_edac module.
1. For Family 19h Models 70h-7Fh all UMCs are not enumerated by the module. 4 are present but only 2 are
enumerated. Currently, required approvals are being sought to upstream the fix. I think this is the issue
being encountered by Sebastian from the dmesg logs and lshw output provided in this thread.
2. For calculating DIMM sizes, the module does not consider both Address Mask and Address Mask Secondary
registers. Consequently, for non-power-of-2 DIMMs, the memory size logged by the module is inaccurate. I
think this is the issue being encountered by Žilvinas.
The patch that Boris asked you folks to try out fixes the second issue. Fix for the first issue has not
yet been submitted.
> I may not have built the kernel correct though... I did not had an environment to do that and just set one up. Or do I need a newer kernel version to apply the patch to? Sorry for the noob questions...
>
> Best,
> Sebastian
--
Thanks,
Avadhut Naik
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: Memory controller not showing half of the memory?
2025-04-01 18:00 ` Naik, Avadhut
@ 2025-07-24 17:47 ` mailinglist
0 siblings, 0 replies; 13+ messages in thread
From: mailinglist @ 2025-07-24 17:47 UTC (permalink / raw)
To: Naik, Avadhut, Žilvinas Žaltiena
Cc: yazen.ghannam, linux-edac, Borislav Petkov
Hi everyone,
I installed 6.12.38 now and - as announced in the 6.12.35 changelog - it
is indeed fixed!
Thanks a lot!
Best,
Sebastian
Am 01.04.2025 um 20:00 schrieb Naik, Avadhut:
> On 4/1/2025 06:44, mailinglist@reox.at wrote:
>>
>>
>> Am 28.03.2025 um 16:24 schrieb Borislav Petkov:
>>> On Thu, Mar 20, 2025 at 11:38:58AM +0200, Žilvinas Žaltiena wrote:
>>>> Hello,
>>>>
>>>> I am also having this "half" memory problem - EDAC is showing 48GB instead
>>>> of 96GB, however otherwise whole 96GB are usable by OS.
>>>
>>> Both of y'all folks encountering this issue, care to run the below patch?
>>>
>>> https://lore.kernel.org/all/20250327210718.1640762-1-avadhut.naik@amd.com/
>>>
>>> Thx.
>>>
>>
>> I tried to apply it on the current Bookworm Backports kernel (6.12.12) - which seemed to work but I still get the same output, i.e., 32GB out of 64GB of RAM.
>>
>
> Think we are mixing up issues here. So, to clarify, there are two issues currently being worked on in the
> amd64_edac module.
>
> 1. For Family 19h Models 70h-7Fh all UMCs are not enumerated by the module. 4 are present but only 2 are
> enumerated. Currently, required approvals are being sought to upstream the fix. I think this is the issue
> being encountered by Sebastian from the dmesg logs and lshw output provided in this thread.
>
> 2. For calculating DIMM sizes, the module does not consider both Address Mask and Address Mask Secondary
> registers. Consequently, for non-power-of-2 DIMMs, the memory size logged by the module is inaccurate. I
> think this is the issue being encountered by Žilvinas.
>
> The patch that Boris asked you folks to try out fixes the second issue. Fix for the first issue has not
> yet been submitted.
>
>> I may not have built the kernel correct though... I did not had an environment to do that and just set one up. Or do I need a newer kernel version to apply the patch to? Sorry for the noob questions...
>>
>> Best,
>> Sebastian
>
^ permalink raw reply [flat|nested] 13+ messages in thread
end of thread, other threads:[~2025-07-24 17:56 UTC | newest]
Thread overview: 13+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2025-03-13 17:21 Memory controller not showing half of the memory? reox
2025-03-13 17:31 ` Borislav Petkov
2025-03-13 17:48 ` reox
2025-03-17 12:07 ` Yazen Ghannam
2025-03-17 12:37 ` reox
2025-03-20 9:38 ` Žilvinas Žaltiena
2025-03-28 15:24 ` Borislav Petkov
2025-03-30 8:25 ` Žilvinas
2025-03-30 10:39 ` Borislav Petkov
2025-03-30 11:00 ` Žilvinas Žaltiena
2025-04-01 11:44 ` mailinglist
2025-04-01 18:00 ` Naik, Avadhut
2025-07-24 17:47 ` mailinglist
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox