linux-kernel - [RFC] de-asmify the x86-64 system call slowpath

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: Windows password security audit tool. GUI, reports in PDF.

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

Date:	Sun, 26 Jan 2014 14:28:15 -0800
From:	Linus Torvalds <torvalds@...ux-foundation.org>
To:	Peter Anvin <hpa@...or.com>, Ingo Molnar <mingo@...hat.com>,
	Thomas Gleixner <tglx@...utronix.de>,
	Peter Zijlstra <peterz@...radead.org>
Cc:	"the arch/x86 maintainers" <x86@...nel.org>,
	Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: [RFC] de-asmify the x86-64 system call slowpath

The x86-64 (and 32-bit, for that matter) system call slowpaths are all
in C, but the *selection* of which slow-path to take is a mixture of
complicated assembler ("sysret_check -> sysret_careful ->
sysret_signal ->sysret_audit -> int_check_syscall_exit_work" etc), and
oddly named and placed C code ("schedule_user" vs
"__audit_syscall_exit" vs "do_notify_resume").

This attached patch tries to take the "do_notify_resume()" approach,
and renaming it to something sane ("syscall_exit_slowpath") and call
out to *all* the different slow cases from that one place, instead of
having some cases hardcoded in asm, and some in C. And instead of
hardcoding which cases result in a "iretq" and which cases result in a
faster sysret case, it's now simply a return value from that
syscall_exit_slowpath() function, so it's very natural and easy to say
"taking a signal will force us to do the slow iretq case, but we can
do the task exit work and still do the sysret".

I've marked this as an RFC, because I didn't bother trying to clean up
the 32-bit code similarly (no test-cases, and trust me, if you get
this wrong, it will fail spectacularly but in very subtle and
hard-to-debug ways), and I also didn't bother with the slow cases in
the "iretq" path, so that path still has the odd asm cases and calls
the old (now legacy) do_notify_resume() path.

But this is actually tested, and seems to work (including limited
testing with strace, gdb etc), and while it adds a few more lines than
it removes, the removed lines are mostly asm, and added lines are C
(and part of them are the temporary still-extant do_notify_resume()
wrapper). In particular, it should be fairly straightforward to take
this as a starting point, removing the extant
do_notify_resume/schedule/etc cases one by one, and get rid of more
asm code and finally the wrapper.

Comments? This was obviously brought on by my frustration with the
currently nasty do_notify_resume() always returning to iret for the
task_work case, and PeterZ's patch that fixed that, but made the asm
mess even *worse*.

                        Linus

View attachment "patch.diff" of type "text/plain" (4573 bytes)