linux-kernel - Re: [PATCH v9 15/16] entry: Inline syscall_exit

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <b382cc8a-3ce9-4fb1-bc0a-a3d9796251d1@arm.com>
Date: Tue, 9 Dec 2025 14:48:51 +0100
From: Kevin Brodsky <kevin.brodsky@....com>
To: Jinjie Ruan <ruanjinjie@...wei.com>, catalin.marinas@....com,
 will@...nel.org, oleg@...hat.com, tglx@...utronix.de, peterz@...radead.org,
 luto@...nel.org, shuah@...nel.org, kees@...nel.org, wad@...omium.org,
 deller@....de, macro@...am.me.uk, charlie@...osinc.com, ldv@...ace.io,
 mark.rutland@....com, song@...nel.org, ryan.roberts@....com,
 ada.coupriediaz@....com, anshuman.khandual@....com, broonie@...nel.org,
 pengcan@...inos.cn, dvyukov@...gle.com,
 linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org,
 linux-kselftest@...r.kernel.org
Subject: Re: [PATCH v9 15/16] entry: Inline syscall_exit_work()

On 04/12/2025 09:21, Jinjie Ruan wrote:
> After switch arm64 to Generic Entry, a new hotspot syscall_exit_work()
> appeared because syscall_exit_work() is no longer inlined. so inline

Before this series the call to syscall_trace_exit() in el0_svc_common()
could not be inlined, so "no longer inlined" doesn't seem to be accurate.

> syscall_exit_work(), and it has 2.6% performance uplift on perf bench
> basic syscall on kunpeng920 as below.

That seems strange. syscall_exit_work() is only called if some flag in
SYSCALL_WORK_EXIT is set, which means that we're doing something special
like tracing. That shouldn't be the case when running a simple perf
bench syscall.

Also worth nothing that its counterpart (syscall_trace_enter())) is not
currently inlined, the asymmetry would have to be justified.

>     | Metric     | W/O this patch | With this patch | Change    |
>     | ---------- | -------------- | --------------- | --------- |
>     | Total time | 2.171 [sec]    | 2.114 [sec]     |  ↓2.6%    |
>     | usecs/op   | 0.217192       | 0.211453        |  ↓2.6%    |
>     | ops/sec    | 4,604,225      | 4,729,178       |  ↑2.7%    |
>
> Signed-off-by: Jinjie Ruan <ruanjinjie@...wei.com>
> ---
>  include/linux/entry-common.h  | 63 ++++++++++++++++++++++++++++++++++-
>  kernel/entry/syscall-common.c | 59 ++------------------------------

These changes are purely generic, surely all architectures using
GENERIC_ENTRY should get similar benefits (assuming LTO isn't used)?

>  2 files changed, 64 insertions(+), 58 deletions(-)
>
> diff --git a/include/linux/entry-common.h b/include/linux/entry-common.h
> index cd6dacb2d8bf..2f84377fb016 100644
> --- a/include/linux/entry-common.h
> +++ b/include/linux/entry-common.h
> @@ -2,6 +2,7 @@
>  #ifndef __LINUX_ENTRYCOMMON_H
>  #define __LINUX_ENTRYCOMMON_H
>  
> +#include <linux/audit.h>
>  #include <linux/irq-entry-common.h>
>  #include <linux/ptrace.h>
>  #include <linux/seccomp.h>
> @@ -128,6 +129,41 @@ static __always_inline long syscall_enter_from_user_mode(struct pt_regs *regs, l
>  	return ret;
>  }
>  
> +/*
> + * If SYSCALL_EMU is set, then the only reason to report is when
> + * SINGLESTEP is set (i.e. PTRACE_SYSEMU_SINGLESTEP).  This syscall
> + * instruction has been already reported in syscall_enter_from_user_mode().
> + */
> +static __always_inline bool report_single_step(unsigned long work)
> +{
> +	if (work & SYSCALL_WORK_SYSCALL_EMU)
> +		return false;
> +
> +	return work & SYSCALL_WORK_SYSCALL_EXIT_TRAP;
> +}
> +
> +/**
> + * arch_ptrace_report_syscall_exit - Architecture specific
> + *				     ptrace_report_syscall_exit.
> + *
> + * Invoked from syscall_exit_work() to wrap ptrace_report_syscall_exit().
> + *
> + * The main purpose is to support arch-specific ptrace_report_syscall_exit
> + * implementation.
> + */
> +static __always_inline void arch_ptrace_report_syscall_exit(struct pt_regs *regs,
> +							    int step);
> +
> +#ifndef arch_ptrace_report_syscall_exit
> +static __always_inline void arch_ptrace_report_syscall_exit(struct pt_regs *regs,
> +							    int step)
> +{
> +	ptrace_report_syscall_exit(regs, step);
> +}
> +#endif

If we want syscall_exit_work() to be inline, then why would we define
this hook in syscall-common.c in patch 12? Might as well define both
hooks in entry-common.h right away and avoid some noise here.

- Kevin

> [...]