[Crash-utility] Re: [PATCH v2 3/3] Enable resolving mangled Rust symbol in lockless ring buffer

Hello Tao, Have you reproduced the issue on s390? Just trigger a segfault in the test program (e.g. by de-referencing a null pointer), run crash on a live system and check the 'log' output.

On 10/7/2025 1:36 PM, Mikhail Zaslonko wrote: > Hi Tao, > > On 10/7/2025 1:11 AM, Tao Liu wrote: >> On Tue, Sep 30, 2025 at 8:18 PM lijiang <lijiang(a)redhat.com&gt; wrote: >>> >>> On Tue, Sep 30, 2025 at 3:01 PM Mikhail Zaslonko <zaslonko(a)linux.ibm.com&gt; wrote: >>>> >>>> Hello, >>>> >>>> On 9/30/2025 4:53 AM, lijiang wrote: >>>>> On Tue, Sep 30, 2025 at 2:18 AM Mikhail Zaslonko <zaslonko(a)linux.ibm.com&gt; >>>>> wrote: >>>>> >>>>>> Hi Lianbo, >>>>>> >>>>>> It seems we have a problem printing disassembly blocks in the log after >>>>>> this patch (at least on s390). >>>>>> See number of empty lines printed below. >>>>>> >>>>> >>>>> Can you try it with the following two patches? >>>>> [1] >>>>> https://www.mail-archive.com/devel@lists.crash-utility.osci.io/msg01621.html >>>> This patch is present on crash master, doesnt affect the issue. >>>>> [2] >>>>> https://www.mail-archive.com/devel@lists.crash-utility.osci.io/msg01636.html >>>>> >>>> This one helps indeed! At least I do not se the blanks lines in my scenario. >>>> BUT, if I use 'log -R' the blank lines are back again! >>> >>> >>> Thanks for the information, Mikhail. >>> Let me see how to debug this(although I can not reproduce it on s390x). >> >> I didn't reproduce the issue on my s390x machine either, maybe sharing >> the vmcore to us can help. > > I'm not sure I'm allowed to send this data outside. > > I think the problem takes place for the logs printed with pr_cont() by the > kernel (e.g. show_code() of arch/s390/kernel/dis.c for printing disassembly as > in my sample). > >> >> Thanks, >> Tao Liu >> >>> >>> Thanks >>> Lianbo >>> >>>> >>>> >>>>> >>>>>> Before the patch: >>>>>> [ 21.408784] Krnl GPRS: 000002aa3780c000 0000000000238c00 >>>>>> 0000000015800000 0000000000001000 >>>>>> [ 21.408787] 0000000008e30000 00000000b0001000 >>>>>> 0000000000000000 0000000000000a55 >>>>>> [ 21.408790] 0000000004c52300 000002aa3780c000 >>>>>> 00000000084281d0 0000038000bfbd28 >>>>>> [ 21.408794] 00000000075e9500 0000000004c52300 >>>>>> 0000038000bfbbf0 0000038000bfbb90 >>>>>> [ 21.408860] Krnl Code: 000000000e0acf48: c05fb0001000 llilf >>>>>> %r5,2952794112 >>>>>> 000000000e0acf4e: ec4100b30659 risbgn >>>>>> %r4,%r1,0,179,6 >>>>>> #000000000e0acf54: 0e24 mvcl >>>>>> %r2,%r4 >>>>>> >000000000e0acf56: a7280000 lhi >>>>>> %r2,0 >>>>>> 000000000e0acf5a: eb6ff0a80004 lmg >>>>>> %r6,%r15,168(%r15) >>>>>> 000000000e0acf60: b9140022 lgfr >>>>>> %r2,%r2 >>>>>> 000000000e0acf64: 07fe bcr >>>>>> 15,%r14 >>>>>> 000000000e0acf66: 47000700 bc >>>>>> 0,1792 >>>>>> [ 21.408883] Call Trace: >>>>>> >>>>>> After the patch: >>>>>> [ 21.408784] Krnl GPRS: 000002aa3780c000 0000000000238c00 >>>>>> 0000000015800000 0000000000001000 >>>>>> [ 21.408787] 0000000008e30000 00000000b0001000 >>>>>> 0000000000000000 0000000000000a55 >>>>>> [ 21.408790] 0000000004c52300 000002aa3780c000 >>>>>> 00000000084281d0 0000038000bfbd28 >>>>>> [ 21.408794] 00000000075e9500 0000000004c52300 >>>>>> 0000038000bfbbf0 0000038000bfbb90 >>>>>> [ 21.408860] >>>>>> >>>>>> >>>>>> >>>>>> >>>>>> >>>>>> >>>>>> Krnl Code: 000000000e0acf48: c05fb0001000 llilf >>>>>> %r5,2952794112 >>>>>> [ 21.408883] Call Trace: >>>>>> >>>>>> I think your approach of storing printable characters in the buffer with >>>>>> the intention to modify it >>>>>> afterwards (see [1] below) instead of writing it character wise does not >>>>>> work when the input data >>>>>> contains several lines. Those '\n' are written one by one, what leads to >>>>>> the result shown above. >>>>>> You should probably buffer and 'demangle' each line separately. >>>>>> >>>>> >>>>> Looks like different issues. Can you share your vmcore and vmlinux with >>>>> me? I did not reproduce it. >>>> >>>> >>>>> >>>>> Thanks >>>>> Lianbo >>>>> >>>>>> >>>>>> >>>>> >>>> >> >