| Message ID | 33059074b78110d4717efe09b887dd28ac77fe7f.camel@pengutronix.de (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | PROBLEM: Segfault in kconfig | expand |
Hi everyone, On Thu, 2022-09-22 at 18:01 +0200, Johannes Zink wrote: > Hi everyone, > > [1.] One line summary of the problem: > kconfig crashes with segfault under rare circumstances > [2.] Full description of the problem/report: > Under certain circumstances jump keys are displayed on the > search > results even if a symbol is deactivated by one of its > dependencies. Using the jump keys then triggers a segmentation > fault due to a NULL dereference. Perform the following steps to > trigger the issue > > 1.: ARCH=arm64 make defconfig > 2.: ARCH=arm64 make menuconfig > > 3.: press '/' key to search for the string "EFI". Use jump key > (1) to jump to search result. Press 'n' key to deactivate > the > entry. > 4.: press '/' to seach for the string "ACPI". Use the jump key > (1) to jump to the search result. > > Menuconfig then crashes with a segfault. > > [3.] Keywords (i.e., modules, networking, kernel): > kconfig, mconf > [4.] Kernel information > [4.1.] Kernel version (from /proc/version): > v6.0.0-rc6 > [4.2.] Kernel .config file: > arm64 default defconfig > [5.] Most recent kernel version which did not have the bug: > v5.15 > [6.] Output of Oops.. message (if applicable) with symbolic > information > resolved (see Documentation/admin-guide/bug-hunting.rst): > not applicable > [7.] A small shell script or example program which triggers the > problem (if possible): > not applicable, please see description in [2.] > [8.] Environment > [8.1.] Software (add the output of the ver_linux script here): > not applicable > [8.2.] Processor information (from /proc/cpuinfo): > not applicable > [8.3.] Module information (from /proc/modules): > not applicable > [8.4.] Loaded driver and hardware information (/proc/ioports, > /proc/iomem): > not applicable > [8.5.] PCI information ('lspci -vvv' as root): > not applicable > [8.6.] SCSI information (from /proc/scsi/scsi): > not applicable > [8.7.] Other information that might be relevant to the problem > (please look in /proc and include all information that you > think to be relevant): > not applicable > [X.] Other notes, patches, fixes, workarounds: > > I found that the attached patch is a very hacky workaround to > keep menuconfig from crashing, but I am pretty sure the jump > keys should not have be activated for unaccessable entries in > the > first place. I found it quite hard to find the corresponding > part > in mconf, which is why I decided to send this bugreport instead > of sending a patch. Maybe someone on this list either knows > mconf > really well and can just fix it, or guide me to where I can dig > around (though in that case I could really use some help on how > to debug menuconfig, since I found it challenging to get it > working with gdb) > > Best regards > Johannes > > --- > scripts/kconfig/mconf.c | 3 ++- > 1 file changed, 2 insertions(+), 1 deletion(-) > > diff --git a/scripts/kconfig/mconf.c b/scripts/kconfig/mconf.c > index 9d3cf510562f..60a82f701bd3 100644 > --- a/scripts/kconfig/mconf.c > +++ b/scripts/kconfig/mconf.c > @@ -447,7 +447,8 @@ static void search_conf(void) > again = false; > for (i = 0; i < JUMP_NB && keys[i]; i++) > if (dres == keys[i]) { > - conf(targets[i]->parent, targets[i]); > + if (targets[i]->parent) > + conf(targets[i]->parent, > targets[i]); > again = true; > } > str_free(&res); > Just a gentle ping and TL;DR on this issue: Menuconfig crashes with a segfault if performing the steps 1.: ARCH=arm64 make defconfig 2.: ARCH=arm64 make menuconfig 3.: press '/' key to search for the string "EFI". Use jump key (1) to jump to search result. Press 'n' key to deactivate the entry. 4.: press '/' to seach for the string "ACPI". Use the jump key (1) to jump to the search result. For more details please see the detailed report in the original message. Has anyone any input on this? Best regards
Hi Johannes, Sorry for the delay. I took a closer look at this. I submitted a patch, which I think is a more correct fix. https://patchwork.kernel.org/project/linux-kbuild/patch/20221023191055.85098-1-masahiroy@kernel.org/ Thanks. On Thu, Oct 6, 2022 at 10:15 PM Johannes Zink <j.zink@pengutronix.de> wrote: > > Hi everyone, > > On Thu, 2022-09-22 at 18:01 +0200, Johannes Zink wrote: > > Hi everyone, > > > > [1.] One line summary of the problem: > > kconfig crashes with segfault under rare circumstances > > [2.] Full description of the problem/report: > > Under certain circumstances jump keys are displayed on the > > search > > results even if a symbol is deactivated by one of its > > dependencies. Using the jump keys then triggers a segmentation > > fault due to a NULL dereference. Perform the following steps to > > trigger the issue > > > > 1.: ARCH=arm64 make defconfig > > 2.: ARCH=arm64 make menuconfig > > > > 3.: press '/' key to search for the string "EFI". Use jump key > > (1) to jump to search result. Press 'n' key to deactivate > > the > > entry. > > 4.: press '/' to seach for the string "ACPI". Use the jump key > > (1) to jump to the search result. > > > > Menuconfig then crashes with a segfault. > > > > [3.] Keywords (i.e., modules, networking, kernel): > > kconfig, mconf > > [4.] Kernel information > > [4.1.] Kernel version (from /proc/version): > > v6.0.0-rc6 > > [4.2.] Kernel .config file: > > arm64 default defconfig > > [5.] Most recent kernel version which did not have the bug: > > v5.15 > > [6.] Output of Oops.. message (if applicable) with symbolic > > information > > resolved (see Documentation/admin-guide/bug-hunting.rst): > > not applicable > > [7.] A small shell script or example program which triggers the > > problem (if possible): > > not applicable, please see description in [2.] > > [8.] Environment > > [8.1.] Software (add the output of the ver_linux script here): > > not applicable > > [8.2.] Processor information (from /proc/cpuinfo): > > not applicable > > [8.3.] Module information (from /proc/modules): > > not applicable > > [8.4.] Loaded driver and hardware information (/proc/ioports, > > /proc/iomem): > > not applicable > > [8.5.] PCI information ('lspci -vvv' as root): > > not applicable > > [8.6.] SCSI information (from /proc/scsi/scsi): > > not applicable > > [8.7.] Other information that might be relevant to the problem > > (please look in /proc and include all information that you > > think to be relevant): > > not applicable > > [X.] Other notes, patches, fixes, workarounds: > > > > I found that the attached patch is a very hacky workaround to > > keep menuconfig from crashing, but I am pretty sure the jump > > keys should not have be activated for unaccessable entries in > > the > > first place. I found it quite hard to find the corresponding > > part > > in mconf, which is why I decided to send this bugreport instead > > of sending a patch. Maybe someone on this list either knows > > mconf > > really well and can just fix it, or guide me to where I can dig > > around (though in that case I could really use some help on how > > to debug menuconfig, since I found it challenging to get it > > working with gdb) > > > > Best regards > > Johannes > > > > --- > > scripts/kconfig/mconf.c | 3 ++- > > 1 file changed, 2 insertions(+), 1 deletion(-) > > > > diff --git a/scripts/kconfig/mconf.c b/scripts/kconfig/mconf.c > > index 9d3cf510562f..60a82f701bd3 100644 > > --- a/scripts/kconfig/mconf.c > > +++ b/scripts/kconfig/mconf.c > > @@ -447,7 +447,8 @@ static void search_conf(void) > > again = false; > > for (i = 0; i < JUMP_NB && keys[i]; i++) > > if (dres == keys[i]) { > > - conf(targets[i]->parent, targets[i]); > > + if (targets[i]->parent) > > + conf(targets[i]->parent, > > targets[i]); > > again = true; > > } > > str_free(&res); > > > > Just a gentle ping and TL;DR on this issue: > > Menuconfig crashes with a segfault if performing the steps > > 1.: ARCH=arm64 make defconfig > 2.: ARCH=arm64 make menuconfig > 3.: press '/' key to search for the string "EFI". Use jump key > (1) to jump to search result. Press 'n' key to deactivate > the entry. > 4.: press '/' to seach for the string "ACPI". Use the jump key > (1) to jump to the search result. > > For more details please see the detailed report in the original > message. > > Has anyone any input on this? > > Best regards > > -- > Pengutronix e.K. | Johannes Zink | > Steuerwalder Str. 21 | https://www.pengutronix.de/ | > 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 | > Amtsgericht Hildesheim, HRA 2686| Fax: +49-5121-206917-5555 | >
Hi Masahiro, thank you for looking into the issue. I will check out your patch as soon as I return from netdevconf next week and test it. Best regards Johannes On Mon, 2022-10-24 at 04:23 +0900, Masahiro Yamada wrote: > Hi Johannes, > > > Sorry for the delay. > > I took a closer look at this. > > > I submitted a patch, which I think is a more correct fix. > > https://patchwork.kernel.org/project/linux-kbuild/patch/20221023191055.85098-1-masahiroy@kernel.org/ > > > > Thanks. > > > On Thu, Oct 6, 2022 at 10:15 PM Johannes Zink <j.zink@pengutronix.de> > wrote: > > > > Hi everyone, > > > > On Thu, 2022-09-22 at 18:01 +0200, Johannes Zink wrote: > > > Hi everyone, > > > > > > [1.] One line summary of the problem: > > > kconfig crashes with segfault under rare circumstances > > > [2.] Full description of the problem/report: > > > Under certain circumstances jump keys are displayed on the > > > search > > > results even if a symbol is deactivated by one of its > > > dependencies. Using the jump keys then triggers a > > > segmentation > > > fault due to a NULL dereference. Perform the following steps > > > to > > > trigger the issue > > > > > > 1.: ARCH=arm64 make defconfig > > > 2.: ARCH=arm64 make menuconfig > > > > > > 3.: press '/' key to search for the string "EFI". Use jump > > > key > > > (1) to jump to search result. Press 'n' key to > > > deactivate > > > the > > > entry. > > > 4.: press '/' to seach for the string "ACPI". Use the jump > > > key > > > (1) to jump to the search result. > > > > > > Menuconfig then crashes with a segfault. > > > > > > [3.] Keywords (i.e., modules, networking, kernel): > > > kconfig, mconf > > > [4.] Kernel information > > > [4.1.] Kernel version (from /proc/version): > > > v6.0.0-rc6 > > > [4.2.] Kernel .config file: > > > arm64 default defconfig > > > [5.] Most recent kernel version which did not have the bug: > > > v5.15 > > > [6.] Output of Oops.. message (if applicable) with symbolic > > > information > > > resolved (see Documentation/admin-guide/bug-hunting.rst): > > > not applicable > > > [7.] A small shell script or example program which triggers the > > > problem (if possible): > > > not applicable, please see description in [2.] > > > [8.] Environment > > > [8.1.] Software (add the output of the ver_linux script here): > > > not applicable > > > [8.2.] Processor information (from /proc/cpuinfo): > > > not applicable > > > [8.3.] Module information (from /proc/modules): > > > not applicable > > > [8.4.] Loaded driver and hardware information (/proc/ioports, > > > /proc/iomem): > > > not applicable > > > [8.5.] PCI information ('lspci -vvv' as root): > > > not applicable > > > [8.6.] SCSI information (from /proc/scsi/scsi): > > > not applicable > > > [8.7.] Other information that might be relevant to the problem > > > (please look in /proc and include all information that you > > > think to be relevant): > > > not applicable > > > [X.] Other notes, patches, fixes, workarounds: > > > > > > I found that the attached patch is a very hacky workaround > > > to > > > keep menuconfig from crashing, but I am pretty sure the jump > > > keys should not have be activated for unaccessable entries > > > in > > > the > > > first place. I found it quite hard to find the corresponding > > > part > > > in mconf, which is why I decided to send this bugreport > > > instead > > > of sending a patch. Maybe someone on this list either knows > > > mconf > > > really well and can just fix it, or guide me to where I can > > > dig > > > around (though in that case I could really use some help on > > > how > > > to debug menuconfig, since I found it challenging to get it > > > working with gdb) > > > > > > Best regards > > > Johannes > > > > > > --- > > > scripts/kconfig/mconf.c | 3 ++- > > > 1 file changed, 2 insertions(+), 1 deletion(-) > > > > > > diff --git a/scripts/kconfig/mconf.c b/scripts/kconfig/mconf.c > > > index 9d3cf510562f..60a82f701bd3 100644 > > > --- a/scripts/kconfig/mconf.c > > > +++ b/scripts/kconfig/mconf.c > > > @@ -447,7 +447,8 @@ static void search_conf(void) > > > again = false; > > > for (i = 0; i < JUMP_NB && keys[i]; i++) > > > if (dres == keys[i]) { > > > - conf(targets[i]->parent, > > > targets[i]); > > > + if (targets[i]->parent) > > > + conf(targets[i]->parent, > > > targets[i]); > > > again = true; > > > } > > > str_free(&res); > > > > > > > Just a gentle ping and TL;DR on this issue: > > > > Menuconfig crashes with a segfault if performing the steps > > > > 1.: ARCH=arm64 make defconfig > > 2.: ARCH=arm64 make menuconfig > > 3.: press '/' key to search for the string "EFI". Use jump key > > (1) to jump to search result. Press 'n' key to deactivate > > the entry. > > 4.: press '/' to seach for the string "ACPI". Use the jump key > > (1) to jump to the search result. > > > > For more details please see the detailed report in the original > > message. > > > > Has anyone any input on this? > > > > Best regards > > > > -- > > Pengutronix e.K. | Johannes Zink | > > Steuerwalder Str. 21 | https://www.pengutronix.de/ | > > 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 | > > Amtsgericht Hildesheim, HRA 2686| Fax: +49-5121-206917-5555 | > > > >
diff --git a/scripts/kconfig/mconf.c b/scripts/kconfig/mconf.c index 9d3cf510562f..60a82f701bd3 100644 --- a/scripts/kconfig/mconf.c +++ b/scripts/kconfig/mconf.c @@ -447,7 +447,8 @@ static void search_conf(void) again = false; for (i = 0; i < JUMP_NB && keys[i]; i++) if (dres == keys[i]) { - conf(targets[i]->parent, targets[i]); + if (targets[i]->parent) + conf(targets[i]->parent, targets[i]); again = true; }
Hi everyone, [1.] One line summary of the problem: kconfig crashes with segfault under rare circumstances [2.] Full description of the problem/report: Under certain circumstances jump keys are displayed on the search results even if a symbol is deactivated by one of its dependencies. Using the jump keys then triggers a segmentation fault due to a NULL dereference. Perform the following steps to trigger the issue 1.: ARCH=arm64 make defconfig 2.: ARCH=arm64 make menuconfig 3.: press '/' key to search for the string "EFI". Use jump key (1) to jump to search result. Press 'n' key to deactivate the entry. 4.: press '/' to seach for the string "ACPI". Use the jump key (1) to jump to the search result. Menuconfig then crashes with a segfault. [3.] Keywords (i.e., modules, networking, kernel): kconfig, mconf [4.] Kernel information [4.1.] Kernel version (from /proc/version): v6.0.0-rc6 [4.2.] Kernel .config file: arm64 default defconfig [5.] Most recent kernel version which did not have the bug: v5.15 [6.] Output of Oops.. message (if applicable) with symbolic information resolved (see Documentation/admin-guide/bug-hunting.rst): not applicable [7.] A small shell script or example program which triggers the problem (if possible): not applicable, please see description in [2.] [8.] Environment [8.1.] Software (add the output of the ver_linux script here): not applicable [8.2.] Processor information (from /proc/cpuinfo): not applicable [8.3.] Module information (from /proc/modules): not applicable [8.4.] Loaded driver and hardware information (/proc/ioports, /proc/iomem): not applicable [8.5.] PCI information ('lspci -vvv' as root): not applicable [8.6.] SCSI information (from /proc/scsi/scsi): not applicable [8.7.] Other information that might be relevant to the problem (please look in /proc and include all information that you think to be relevant): not applicable [X.] Other notes, patches, fixes, workarounds: I found that the attached patch is a very hacky workaround to keep menuconfig from crashing, but I am pretty sure the jump keys should not have be activated for unaccessable entries in the first place. I found it quite hard to find the corresponding part in mconf, which is why I decided to send this bugreport instead of sending a patch. Maybe someone on this list either knows mconf really well and can just fix it, or guide me to where I can dig around (though in that case I could really use some help on how to debug menuconfig, since I found it challenging to get it working with gdb) Best regards Johannes --- scripts/kconfig/mconf.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) str_free(&res);