Skip to contentSkip to conversation

RAS/CEC: Move non-debug attributes out of debugfs

PATCH In review kyle-meyer 4 participants 7 messages Mar 23, 2023 – Mar 27, 2023

Subsystems (4)
lore
Share & tools
lore

kyle-meyer

author
#0 lore raw

PATCH 1/1 RAS/CEC: Move non-debug attributes out of debugfs

From: Kyle Meyer <[email protected]>

When kernel lockdown is in effect, use of debugfs is not permitted. Move decay_interval and action_threshold out of debugfs, from debugfs/ras/cec to sysfs/system/devices/machinecheck/cec.

---
 arch/x86/include/asm/mce.h     |   1 +
 arch/x86/kernel/cpu/mce/core.c |   3 +-
 arch/x86/ras/Kconfig           |   4 +-
 drivers/ras/cec.c              | 141 ++++++++++++++++++++++-----------
 4 files changed, 101 insertions(+), 48 deletions(-)
arch/x86/include/asm/mce.h +1 -0
index 9646ed6e8c0b..c5d358499899 100644
@@ -206,6 +206,7 @@ static inline void enable_copy_mc_fragile(void)
 struct cper_ia_proc_ctx;
 
 #ifdef CONFIG_X86_MCE
+extern struct bus_type mce_subsys;
 int mcheck_init(void);
 void mcheck_cpu_init(struct cpuinfo_x86 *c);
 void mcheck_cpu_clear(struct cpuinfo_x86 *c);
arch/x86/kernel/cpu/mce/core.c +2 -1
index 2eec60f50057..1a3eaa501ae4 100644
@@ -2376,10 +2376,11 @@ static void mce_enable_ce(void *all)
 		__mcheck_cpu_init_timer();
 }
 
-static struct bus_type mce_subsys = {
+struct bus_type mce_subsys = {
 	.name		= "machinecheck",
 	.dev_name	= "machinecheck",
 };
+EXPORT_SYMBOL_GPL(mce_subsys);
 
 DEFINE_PER_CPU(struct device *, mce_device);
 
arch/x86/ras/Kconfig +2 -2
index 7488c715427e..5f5f6f9a5f3c 100644
@@ -1,7 +1,7 @@
 # SPDX-License-Identifier: GPL-2.0
 config RAS_CEC
 	bool "Correctable Errors Collector"
-	depends on X86_MCE && MEMORY_FAILURE && DEBUG_FS
+	depends on X86_MCE && MEMORY_FAILURE
 	help
 	  This is a small cache which collects correctable memory errors per 4K
 	  page PFN and counts their repeated occurrence. Once the counter for a
@@ -15,7 +15,7 @@ config RAS_CEC
 config RAS_CEC_DEBUG
 	bool "CEC debugging machinery"
 	default n
-	depends on RAS_CEC
+	depends on RAS_CEC && DEBUG_FS
 	help
 	  Add extra files to (debugfs)/ras/cec to test the correctable error
 	  collector feature. "pfn" is a writable file that allows user to
drivers/ras/cec.c +96 -45

206 lines not on this page · raw message

Signature · 1 line
2.26.2
#1 lore raw
Quoted 5 lines · kyle-meyer

From: Kyle Meyer <[email protected]>

When kernel lockdown is in effect, use of debugfs is not permitted. Move decay_interval and action_threshold out of debugfs, from debugfs/ras/cec to sysfs/system/devices/machinecheck/cec.

You forgot to attach the patch that you wrote that updates Documentation/ABI/. ;)

Also, why *should* these be part of the stable sysfs ABI? What app is using them? Why does it need them? Why these two and only these two? What's left in debugfs?

#2 lore raw
Quoted 2 lines

You forgot to attach the patch that you wrote that updates Documentation/ABI/. ;)

Thanks, I will include the documentation with a second version of the patch.

Quoted 2 lines

Also, why *should* these be part of the stable sysfs ABI? What app is using them? Why does it need them?

We have system scripts that adjust decay_interval and action_threshold. They can't access those attributes when lockdown is enabled. If there is a more appropriate place for those attributes, please let me know.

Quoted 1 line

Why these two and only these two? What's left in debugfs?

The other attributes (pfn and array) are used to test CEC. They are only created when RAS_CEC_DEBUG is enabled.

Thanks,
Kyle Meyer

#3 lore raw
Quoted 5 lines · Kyle
Quoted 2 lines

Also, why *should* these be part of the stable sysfs ABI? What app is using them? Why does it need them?

We have system scripts that adjust decay_interval and action_threshold. They can't access those attributes when lockdown is enabled. If there is a more appropriate place for those attributes, please let me know.

Thanks for the info. That helps a bit. But, I'd also appreciate if you could expand on this a little more. What "system scripts" are these? Who is using them? What are they trying to accomplish?

We can try to find the best home for these attributes with that info in hand, if it's not sysfs.

Quoted 3 lines
Quoted 1 line

Why these two and only these two? What's left in debugfs?

The other attributes (pfn and array) are used to test CEC. They are only created when RAS_CEC_DEBUG is enabled.

Oh, that's good info too. Can you please include that in some form in your new changelog?

maintainer
#4 lore raw
In reply to kyle-meyer
Quoted 5 lines · kyle-meyer

From: Kyle Meyer <[email protected]>

When kernel lockdown is in effect, use of debugfs is not permitted. Move decay_interval and action_threshold out of debugfs, from debugfs/ras/cec to sysfs/system/devices/machinecheck/cec.

All those knobs are in debugfs because we wanted to discuss the proper interface design first and only then cast them in stone. I guess that has not happened yet.

What you're doing is certainly not what we had in mind so just because some lockdown policy says so, is not good enough.

Quoted 5 lines
arch/x86/kernel/cpu/mce/core.c +0 -0
index 2eec60f50057..1a3eaa501ae4 100644

arch/x86/kernel/cpu/mce/core.c static void mce_enable_ce(void *all) · line 2383

Quoted 9 lines
arch/x86/kernel/cpu/mce/core.c +2 -1
@@ -2376,10 +2376,11 @@ static void mce_enable_ce(void *all)
 		__mcheck_cpu_init_timer();
 }
 
-static struct bus_type mce_subsys = {
+struct bus_type mce_subsys = {
 	.name		= "machinecheck",
 	.dev_name	= "machinecheck",
 };
+EXPORT_SYMBOL_GPL(mce_subsys);

Nope, this is not going to happen.

Besides, that error collector is used on x86 but it is generic enough so that it can be used by other arches. So if anything, it should not be exposed in the x86's "machinecheck" hierarchy but somewhere generic.

And until that proper interface has been hammered out, you can just as well disable it in your lockdown configs.

Thx.

Signature · 4 lines
Regards/Gruss,
    Boris.

https://people.kernel.org/tglx/notes-about-netiquette
#5 lore raw

On Fri, Mar 24, 2023 at 01:49:19AM +0100, Borislav Petkov wrote:
...

Quoted 5 lines

Besides, that error collector is used on x86 but it is generic enough so that it can be used by other arches. So if anything, it should not be exposed in the x86's "machinecheck" hierarchy but somewhere generic.

How about "/sys/ras" to collect global RAS interfaces? Maybe it can link to other directories like "/sys/devices/*" for hardware things like MCA, and "/sys/kernel/*" for kernel things like CEC.

Thanks,
Yazen

maintainer
#6 lore raw
Quoted 1 line · Yazen Ghannam

How about "/sys/ras" to collect global RAS interfaces?

Looking a bit at

Documentation/admin-guide/sysfs-rules.rst

it does sound like /sys/<subsystem> could be a good place. Might wanna run it by Greg, though, first.

Quoted 2 lines

Maybe it can link to other directories like "/sys/devices/*" for hardware things like MCA,

Those are at

/sys/devices/system/machinecheck/

Quoted 1 line

and "/sys/kernel/*" for kernel things like CEC.

Or /sys/drivers

Remember there's also

/sys/devices/system/edac/

so a lot of RAS junk spread all over the place.

Signature · 4 lines
Regards/Gruss,
    Boris.

https://people.kernel.org/tglx/notes-about-netiquette