Re: [PATCH] arm64: Make CONFIG_ARM64_PSEUDO_NMI macro wrap all the pseudo-NMI code

From: He Ying
Date: Sun Jan 09 2022 - 22:00:48 EST


Hi Mark,

I'm just back from the weekend and sorry for the delayed reply.


在 2022/1/7 21:19, Mark Rutland 写道:
On Fri, Jan 07, 2022 at 03:55:36AM -0500, He Ying wrote:
Our product has been updating its kernel from 4.4 to 5.10 recently and
found a performance issue. We do a bussiness test called ARP test, which
tests the latency for a ping-pong packets traffic with a certain payload.
The result is as following.

- 4.4 kernel: avg = ~20s
- 5.10 kernel (CONFIG_ARM64_PSEUDO_NMI is not set): avg = ~40s
Have you tested with a recent mainline kernel, e.g. v5.15?

Actuallly no, that's because this test is only available for the product environment and

we don't have an available 5.15 kernel for it yet.


Is this test publicly available, and can you say which hardrware (e.g. which
CPU implementation) you're testing with?

Actually no. The test is only available for our product environment now. We are testing

with hisilicon 1213 (4 ARM Cortex-A72 cores).


I have been just learning arm64 pseudo-NMI code and have a question,
why is the related code not wrapped by CONFIG_ARM64_PSEUDO_NMI?
The code in question is all patched via alternatives, and when
CONFIG_ARM64_PSEUDO_NMI is not selected, the code was expected to only have the
overhead of the regular DAIF manipulation.
I don't understand alernatives very well and I'll apreciate it if you can explain it a bit more.

I wonder if this brings some performance regression.

First, I make this patch and then do the test again. Here's the result.

- 5.10 kernel with this patch not applied: avg = ~40s
- 5.10 kernel with this patch applied: avg = ~23s

Amazing! Note that all kernel is built with CONFIG_ARM64_PSEUDO_NMI not
set. It seems the pseudo-NMI feature actually brings some overhead to
performance event if CONFIG_ARM64_PSEUDO_NMI is not set.
I'm surprised the overhead is so significant; as above this is all patched in
and so the overhead when this is disabled is expected to be *extremely* small.

For example, wjen CONFIG_ARM64_PSEUDO_NMI, in arch_local_irq_enable():

* The portion under the system_has_prio_mask_debugging() test will be removed
entirely by the compiler, as this internally checks
IS_ENABLED(CONFIG_ARM64_PSEUDO_NMI).

* The assembly will be left as a write to DAIFClr. The only additional cost
should be that of generating GIC_PRIO_IRQON into a register.

* The pmr_sync() will be removed entirely by the compiler as is defined
conditionally dependent on CONFIG_ARM64_PSEUDO_NMI.

I can't spot an obvious issue with that or ther other cases. In the common case
those add no new instructions, and in the worst case they only add NOPs.

Thanks for your detailed explaination! Actually I can't understand the result exactly.

I build two 5.10 kernel images with this patch applied or not and objdump them. Indeed,

the disassembles of 'arch_local_irq_restore' are the same. Do you have any ideas how

we can find the root cause why this patch improves the performance so much?


However, the test result is trustworthy because we do it many times and the result is

always repeatable.


Furthermore, I find the feature also brings some overhead to vmlinux size.
I build 5.10 kernel with this patch applied or not while
CONFIG_ARM64_PSEUDO_NMI is not set.

- 5.10 kernel with this patch not applied: vmlinux size is 384060600 Bytes.
- 5.10 kernel with this patch applied: vmlinux size is 383842936 Bytes.

That means arm64 pseudo-NMI feature may bring ~200KB overhead to
vmlinux size.
I suspect that's just the (unused) alternatives, and we could improve that by
passing the config into the alternative blocks.

Do you mean the sections generated by the alternatives? I don't understand

alernatives very well and I'll apreciate it if you can explain it a bit more.


Above all, arm64 pseudo-NMI feature brings some overhead to vmlinux size
and performance even if config is not set. To avoid it, add macro control
all around the related code.

Signed-off-by: He Ying <heying24@xxxxxxxxxx>
---
arch/arm64/include/asm/irqflags.h | 38 +++++++++++++++++++++++++++++--
arch/arm64/kernel/entry.S | 4 ++++
2 files changed, 40 insertions(+), 2 deletions(-)

diff --git a/arch/arm64/include/asm/irqflags.h b/arch/arm64/include/asm/irqflags.h
index b57b9b1e4344..82f771b41cf5 100644
--- a/arch/arm64/include/asm/irqflags.h
+++ b/arch/arm64/include/asm/irqflags.h
@@ -26,6 +26,7 @@
*/
static inline void arch_local_irq_enable(void)
{
+#ifdef CONFIG_ARM64_PSEUDO_NMI
if (system_has_prio_mask_debugging()) {
u32 pmr = read_sysreg_s(SYS_ICC_PMR_EL1);
@@ -41,10 +42,18 @@ static inline void arch_local_irq_enable(void)
: "memory");
pmr_sync();
+#else
+ asm volatile(
+ "msr daifclr, #3 // arch_local_irq_enable"
+ :
+ :
+ : "memory");
+#endif
I'm happy to rework this to improve matters, but I am very much not happy with
duplicating the logic for the !PSEUDO_NMI case. Adding more ifdeffery and
copies of that is not acceptable.
I agree. Adding these ifdeffery is a bit ugly. Let's see if there are some better ways.

Instead, can you please try changing the alternative to also take the config,
e.g. here have:

| asm volatile(ALTERNATIVE(
| "msr daifclr, #3 // arch_local_irq_enable",
| __msr_s(SYS_ICC_PMR_EL1, "%0"),
| ARM64_HAS_IRQ_PRIO_MASKING,
| CONFIG_ARM64_PSEUDO_NMI)
| :
| : "r" ((unsigned long) GIC_PRIO_IRQON)
| : "memory");

... and see if that makes a significant difference?

Likewise for the other casees.

OK, I'll try it. But I have some questions. Here's the comment of ALERNATIVE:

/*
 * Usage: asm(ALTERNATIVE(oldinstr, newinstr, feature));
 *
 * Usage: asm(ALTERNATIVE(oldinstr, newinstr, feature, CONFIG_FOO));
 * N.B. If CONFIG_FOO is specified, but not selected, the whole block
 *      will be omitted, including oldinstr.
 */

If CONFIG_FOO is not selected, the whole block will be omitted including oldinstr.

But we still want the oldinstr in this situation. Do I misunderstand something?


#endif /* __ASM_IRQFLAGS_H */
diff --git a/arch/arm64/kernel/entry.S b/arch/arm64/kernel/entry.S
index 2f69ae43941d..ffc32d3d909a 100644
--- a/arch/arm64/kernel/entry.S
+++ b/arch/arm64/kernel/entry.S
@@ -300,6 +300,7 @@ alternative_else_nop_endif
str w21, [sp, #S_SYSCALLNO]
.endif
+#ifdef CONFIG_ARM64_PSEUDO_NMI
/* Save pmr */
alternative_if ARM64_HAS_IRQ_PRIO_MASKING
mrs_s x20, SYS_ICC_PMR_EL1
@@ -307,6 +308,7 @@ alternative_if ARM64_HAS_IRQ_PRIO_MASKING
mov x20, #GIC_PRIO_IRQON | GIC_PRIO_PSR_I_SET
msr_s SYS_ICC_PMR_EL1, x20
alternative_else_nop_endif
+#endif
/* Re-enable tag checking (TCO set on exception entry) */
#ifdef CONFIG_ARM64_MTE
@@ -330,6 +332,7 @@ alternative_else_nop_endif
disable_daif
.endif
+#ifdef CONFIG_ARM64_PSEUDO_NMI
/* Restore pmr */
alternative_if ARM64_HAS_IRQ_PRIO_MASKING
ldr x20, [sp, #S_PMR_SAVE]
@@ -339,6 +342,7 @@ alternative_if ARM64_HAS_IRQ_PRIO_MASKING
dsb sy // Ensure priority change is seen by redistributor
.L__skip_pmr_sync\@:
alternative_else_nop_endif
+#endif
For these two I think the ifdeffery is fine, but I'm surprised this has a
measureable impact as the alternatives should be initialized to NOPS (and never
modified).

Yes, these NOPs may bring some overhead. But I can't say how much these NOPs contribute to

the result of test.


Thanks,
Mark.
.