linux-kernel - Re: [PATCH] arm64: scripts/sorttable: Implement sorting mcount

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

Message-ID: <20250213150934.12012-C-hca@linux.ibm.com>
Date: Thu, 13 Feb 2025 16:09:34 +0100
From: Heiko Carstens <hca@...ux.ibm.com>
To: Steven Rostedt <rostedt@...dmis.org>
Cc: Catalin Marinas <catalin.marinas@....com>, Will Deacon <will@...nel.org>,
        LKML <linux-kernel@...r.kernel.org>,
        linux-arm-kernel@...ts.infradead.org,
        Mark Rutland <mark.rutland@....com>,
        Linux Trace Kernel <linux-trace-kernel@...r.kernel.org>,
        Vasily Gorbik <gor@...ux.ibm.com>,
        Alexander Gordeev <agordeev@...ux.ibm.com>, linux-s390@...r.kernel.org
Subject: Re: [PATCH] arm64: scripts/sorttable: Implement sorting mcount_loc
 at boot for arm64

On Thu, Feb 13, 2025 at 01:53:02PM +0100, Heiko Carstens wrote:
> On Mon, Feb 10, 2025 at 02:26:47PM -0500, Steven Rostedt wrote:
> > For the s390 folks. I cross compiled a s390 and looked at the mcount_loc
> > section, and I have no idea how to implement this for that. I wrote a elf
> > parser to dump sections based symbols:
> > 
> >   https://rostedt.org/code/dump_elf_sym.c
> > 
> > And ran it on the s390 vmlinux:
> > 
> >  $ ./dump_elf_sym vmlinux __start_mcount_loc __stop_mcount_loc
> >  1801620: .. .. .. .. .. .. .. ..   00 00 00 00 00 11 70 20  ......... .....p 
> >  1801630: 00 00 00 00 00 11 70 90   00 00 00 00 00 11 70 a0  ......p.. .....p.
> >  1801640: 00 00 00 00 00 11 71 10   00 00 00 00 00 11 71 20  ......q.. .....q 
> >  1801650: 00 00 00 00 00 11 71 90   00 00 00 00 01 7c 70 00  ......q.. ....|p.
> >  1801660: 00 00 00 00 01 7c 70 20   00 00 00 00 01 7c 70 40  .....|p . ....|p@
> >  1801670: 00 00 00 00 01 7c 70 60   00 00 00 00 01 7c 70 70  .....|p`. ....|pp
> >  1801680: 00 00 00 00 01 7c 70 98   00 00 00 00 01 7c 70 c0  .....|p.. ....|p.
> >  1801690: 00 00 00 00 01 7c 70 d0   00 00 00 00 01 7c 71 68  .....|p.. ....|qh
> > [..]
> > 
> > It looks like addresses in that section...
> 
> Those are the addresses of the mcount locations. After looking at
> sorttable.c it really looks like that for s390 we can simply select
> HAVE_BUILDTIME_MCOUNT_SORT without any further changes.
> 
> I just tested it with different compiler options (fentry vs hotpatch),
> including selecting FTRACE_SORT_STARTUP_TEST, and as expected everything
> works.
> 
> I'm going to give it some more testing in our CI - but if nothing breaks a
> patch which selects HAVE_BUILDTIME_MCOUNT_SORT for s390 will go upstream
> with the next merge window.

Something like this:

>From 0759d6b961946b7e5cfb19971d56f5493204301c Mon Sep 17 00:00:00 2001
From: Heiko Carstens <hca@...ux.ibm.com>
Date: Thu, 13 Feb 2025 13:57:33 +0100
Subject: [PATCH] s390: Sort mcount locations at build time

For s390 the mcount_loc section of the kernel image contains the addresses
of the mcount locations. All addresses will be adjusted with the same
offset by the decompressor before the kernel is started.

Therefore select HAVE_BUILDTIME_MCOUNT_SORT so that the entries of this
section are sorted at build time. Given that the same offset is applied to
all entries the section will be sorted in any case.

Note that this was not possible before commit 778666df60f0 ("s390: compile
relocatable kernel without -fPIE"). Since this commit all R_390_64 absolute
relocations are handled in a special way: only the address of the to be
changed location is put into a special section. For all those locations the
same offset is applied as described above.

Without that change it would have been necessary to also adjust the addend
of all relocations which correspond to the mcount_loc section, when sorting
the mcount_loc section.

Reported-by: Steven Rostedt <rostedt@...dmis.org>
Closes: https://lore.kernel.org/r/20250210142647.083ff456@gandalf.local.home/
Signed-off-by: Heiko Carstens <hca@...ux.ibm.com>
---
 arch/s390/Kconfig                 | 1 +
 arch/s390/configs/debug_defconfig | 1 +
 2 files changed, 2 insertions(+)

diff --git a/arch/s390/Kconfig b/arch/s390/Kconfig
index 9c9ec08d78c7..acaa1d1c12b2 100644
--- a/arch/s390/Kconfig
+++ b/arch/s390/Kconfig
@@ -185,6 +185,7 @@ config S390
 	select HAVE_ARCH_TRANSPARENT_HUGEPAGE
 	select HAVE_ARCH_VMAP_STACK
 	select HAVE_ASM_MODVERSIONS
+	select HAVE_BUILDTIME_MCOUNT_SORT
 	select HAVE_CMPXCHG_DOUBLE
 	select HAVE_CMPXCHG_LOCAL
 	select HAVE_DEBUG_KMEMLEAK
diff --git a/arch/s390/configs/debug_defconfig b/arch/s390/configs/debug_defconfig
index d6beec5292a0..a2b0444b7d6b 100644
--- a/arch/s390/configs/debug_defconfig
+++ b/arch/s390/configs/debug_defconfig
@@ -887,6 +887,7 @@ CONFIG_USER_EVENTS=y
 CONFIG_HIST_TRIGGERS=y
 CONFIG_FTRACE_STARTUP_TEST=y
 # CONFIG_EVENT_TRACE_STARTUP_TEST is not set
+CONFIG_FTRACE_SORT_STARTUP_TEST=y
 CONFIG_SAMPLES=y
 CONFIG_SAMPLE_TRACE_PRINTK=m
 CONFIG_SAMPLE_FTRACE_DIRECT=m
-- 
2.45.2