retitle 261958 hotplug: Loads framebuffer without questioning, breaking working system on 2.4 kernel
thanks
This is because pciutils adds pcimodules on 1:2.1.11-13, so hotplug
works pci coldplugging correctly on 2.4 kernel. Unfortunately, some
devices, especially framebuffer devices, won't work fine.
For 2.6 kernel, hotplug/ignore_pci_class_display would be used to
control whether display class pci devices will be handled in hotplug
or not, but for 2.4 kernel, all of pci modules will be handled now.
I'm thinking that I'll add the code in hotplug.functions:load_drivers()
when hotplug/ignore_pci_class_display is true, check module path includes
"drivers/video".
+ # XXX: check PCI video modules
+ # It may break working systems. Bug#261958
+ if [ "$IGNORE_PCI_CLASS_DISPLAY" = "true" ] && $MODPROBE -n -v "$MODULE" | grep -q '/drivers/video/' > /dev/null 2>&1; then
+
+ mesg " $MODULE: ignore display modules"
+ continue
+ fi
+
# statically linked modules aren't shown by 'lsmod',
# and user mode drivers will ONLY have a setup script;
# it's not an error if a module doesn't exist or won't load.
Regards,
Fumitoshi UKAI
At Thu, 29 Jul 2004 09:32:30 +0200,
Helge Kreutzmann wrote:
> After upgrading pciutils from 11-11 to 11-15 hotplug loads the
> framebuffer modules and initializes framebuffer. Since nothing is
> prepared for it, the first console has a broken font (but this
> is also on my ibook), X barly works
> (but no keyboard), and leaving X gives me a almost black "console",
> ie. input only.(2)
>
> I never used framebuffer on this machine and I currently do not intend
> to. And I definitly never consented to loading the appropriate
> modules(1) (although they are build, because at some stage I wanted to
> experiment but did not have time).
>
> All other hotplug systems are harmless, since the modules can be
> unloaded and everything is fine again, this is not the case for
> framebuffer. Also since when is a video device "hotpluggable"?
>
> If this bug is related to pciutils or discover, please reassign. If
> you need more information, please tell me which.
>
> (1) Looking at the debconf-info below this is confirmed (If I
> understand them correctly). And I do not have a USB keyboard.
>
> (2) Yes, these are bugs which should be reported separately and
> independent of hotplug. But hotplug should not change the working
> of a system in such a fundamental way as to open a can of new
> bugs in other software, especially since the users might need the
> machine in the mean time.
>
> -- System Information:
> Debian Release: testing/unstable
> APT prefers testing
> APT policy: (990, 'testing')
> Architecture: alpha
> Kernel: Linux 2.4.26
> Locale: LANG=de_DE@euro, LC_CTYPE=de_DE@euro
>
> Versions of packages hotplug depends on:
> ii debconf 1.4.29 Debian configuration management sy
> ii modutils 2.4.26-1 Linux module utilities
> ii procps 1:3.2.1-2 The /proc file system utilities
>
> -- debconf information:
> hotplug/ignore_pci_class_display: true
> hotplug/net_agent_policy: hotplug
> hotplug/static_module_list:
> hotplug/usbd_enable: true
> * hotplug/usb_keyboard:
> hotplug/x11_usbmice_hack: false
> --
> Helge Kreutzmann, Dipl.-Phys. <email address hidden>
> gpg signed mail preferred gpg-key: finger <email address hidden>
> 64bit GNU powered http://www.itp.uni-hannover.de/~kreutzm
> Help keep free software "libre": http://www.freepatents.org/
> [2 <application/pgp-signature (7bit)>]
>
retitle 261958 hotplug: Loads framebuffer without questioning, breaking working system on 2.4 kernel
thanks
This is because pciutils adds pcimodules on 1:2.1.11-13, so hotplug ignore_ pci_class_ display would be used to
works pci coldplugging correctly on 2.4 kernel. Unfortunately, some
devices, especially framebuffer devices, won't work fine.
For 2.6 kernel, hotplug/
control whether display class pci devices will be handled in hotplug
or not, but for 2.4 kernel, all of pci modules will be handled now.
I'm thinking that I'll add the code in hotplug. functions: load_drivers( ) ignore_ pci_class_ display is true, check module path includes
when hotplug/
"drivers/video".
Does this patch fix your problem?
--- hotplug. functions. orig 2004-07-29 17:21:16.000000000 +0900
+++ hotplug.functions 2004-07-29 17:24:00.000000000 +0900
@@ -160,6 +160,14 @@
continue
fi
+ # XXX: check PCI video modules PCI_CLASS_ DISPLAY" = "true" ] && $MODPROBE -n -v "$MODULE" | grep -q '/drivers/video/' > /dev/null 2>&1; then
+ # It may break working systems. Bug#261958
+ if [ "$IGNORE_
+
+ mesg " $MODULE: ignore display modules"
+ continue
+ fi
+
# statically linked modules aren't shown by 'lsmod',
# and user mode drivers will ONLY have a setup script;
# it's not an error if a module doesn't exist or won't load.
Regards,
Fumitoshi UKAI
At Thu, 29 Jul 2004 09:32:30 +0200,
Helge Kreutzmann wrote:
> After upgrading pciutils from 11-11 to 11-15 hotplug loads the ignore_ pci_class_ display: true net_agent_ policy: hotplug static_ module_ list: usbd_enable: true usb_keyboard: x11_usbmice_ hack: false www.itp. uni-hannover. de/~kreutzm www.freepatents .org/ pgp-signature (7bit)>]
> framebuffer modules and initializes framebuffer. Since nothing is
> prepared for it, the first console has a broken font (but this
> is also on my ibook), X barly works
> (but no keyboard), and leaving X gives me a almost black "console",
> ie. input only.(2)
>
> I never used framebuffer on this machine and I currently do not intend
> to. And I definitly never consented to loading the appropriate
> modules(1) (although they are build, because at some stage I wanted to
> experiment but did not have time).
>
> All other hotplug systems are harmless, since the modules can be
> unloaded and everything is fine again, this is not the case for
> framebuffer. Also since when is a video device "hotpluggable"?
>
> If this bug is related to pciutils or discover, please reassign. If
> you need more information, please tell me which.
>
> (1) Looking at the debconf-info below this is confirmed (If I
> understand them correctly). And I do not have a USB keyboard.
>
> (2) Yes, these are bugs which should be reported separately and
> independent of hotplug. But hotplug should not change the working
> of a system in such a fundamental way as to open a can of new
> bugs in other software, especially since the users might need the
> machine in the mean time.
>
> -- System Information:
> Debian Release: testing/unstable
> APT prefers testing
> APT policy: (990, 'testing')
> Architecture: alpha
> Kernel: Linux 2.4.26
> Locale: LANG=de_DE@euro, LC_CTYPE=de_DE@euro
>
> Versions of packages hotplug depends on:
> ii debconf 1.4.29 Debian configuration management sy
> ii modutils 2.4.26-1 Linux module utilities
> ii procps 1:3.2.1-2 The /proc file system utilities
>
> -- debconf information:
> hotplug/
> hotplug/
> hotplug/
> hotplug/
> * hotplug/
> hotplug/
> --
> Helge Kreutzmann, Dipl.-Phys. <email address hidden>
> gpg signed mail preferred gpg-key: finger <email address hidden>
> 64bit GNU powered http://
> Help keep free software "libre": http://
> [2 <application/
>