[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <6c910206-2409-4a7e-9d5e-8ffc1c269c08@foss.arm.com>
Date: Mon, 20 May 2024 10:42:01 +0100
From: Leo Yan <niayan01@...s.arm.com>
To: Ian Rogers <irogers@...gle.com>
Cc: Arnaldo Carvalho de Melo <acme@...nel.org>,
Adrian Hunter <adrian.hunter@...el.com>, James Clark <james.clark@....com>,
Peter Zijlstra <peterz@...radead.org>, Ingo Molnar <mingo@...hat.com>,
Namhyung Kim <namhyung@...nel.org>, Mark Rutland <mark.rutland@....com>,
Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
Jiri Olsa <jolsa@...nel.org>, Athira Rajeev <atrajeev@...ux.vnet.ibm.com>,
linux-perf-users@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2 0/2] perf maps: Improve the kcore maps merging
On 5/15/24 22:39, Ian Rogers wrote:
[...]
> Thanks Leo, testing this change on perf-tools-next with an x86 debian
> laptop I see:
> ```
> $ perf test 24 -v
> 24: Object code reading:
> --- start ---
> test child forked, pid 3407499
> Looking at the vmlinux_path (8 entries long)
> symsrc__init: build id mismatch for vmlinux.
> symsrc__init: cannot get elf header.
> overlapping maps in [kernel.kallsyms] (disable tui for more info)
> Using /proc/kcore for kernel data
> Using /proc/kallsyms for symbols
> Parsing event 'cycles'
> Using CPUID GenuineIntel-6-8D-1
> mmap size 528384B
> Reading object code for memory address: 0xfffffffface8d64a
> File is: /proc/kcore
> On file address is: 0xfffffffface8d64a
> dso__data_read_offset failed
> ---- end(-1) ----
> 24: Object code reading : FAILED!
> ```
> The test passes without the changes. Let me know if you need me to dig deeper.
Just update for the test failure.
The failure is caused by the memory map's attribute is not updated after
I removed the map replacement related code in patch 02. This means the
old (identical) kernel map (generated by '/proc/kallsyms') still keeps
some stale info for 'pgoff', 'mapping_type', etc.
To fix this issue, in the new patch, it removes the old kernel map (to
avoid any stale info) and update 'machine->vmlinux_map'.
Just one concern. I saw the kcore exposes kernel text section which is
dependent on the CONFIG_ARCH_PROC_KCORE_TEXT config, but some arches
(e.g. PowerPC, etc) are absent for this config. But you can see even the
current code doesn't handle this case, if the replacement map is not
found, it is arbitrarily assigned to the first map in kcore [1]. An
option is to rollback to use the old identical kernel map, this is not
addressed in my latest patch, which is simply report failure in this
case. Please review it and let me know how you think.
Thanks,
Leo
[1]
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/tools/perf/util/symbol.c?h=v6.9#n1366
Powered by blists - more mailing lists