Re: do_softirq solution pending?
"Linuxant support (Jonathan)" <[email protected]> Mon, 25 Feb 2008 16:34:26 -0500
| Newsgroups | gmane.linux.kernel.driver-loader |
|---|---|
| Message-ID | <[email protected]> |
Hi, we are aware that DriverLoader 2.40 does not work on a 2.6.24.x kernel because the "do_softirq" symbol is no longer exported to modules. This issue is difficult to solve in DriverLoader but our developers are working on this. Patching the kernel to export the symbol, like you did Bob, is probably the most straightforward workaround at the moment but should only be attempted by users which know what they are doing. Patching the kernel is not an ideal solution, hopefully something can be done in DriverLoader itself to solve this problem. Regards, Jonathan Technical specialist / Linuxant www.linuxant.com [email protected] Bob Tracy wrote: > rct wrote: >> Can the workaround for the case where CONFIG_4KSTACKS is not defined >> really be as simple as >> >> #define irq_exit() >> >> in "modules/GPL/oscompat.h"? > > Never mind: I'm a bozo. Got into interrupt processing trouble in a > hurry, and when I shutdown driverloader to stop the continuous updates > to syslog, the machine locked up solid. > > On the other hand, setting CONFIG_4KSTACKS and applying the following > small patch (against 2.6.24-rc6) *does* work: > > --- linux/arch/x86/kernel/irq_32.c.orig 2007-11-08 08:05:14.000000000 -0600 > +++ linux/arch/x86/kernel/irq_32.c 2008-02-24 17:13:34.000000000 -0600 > @@ -229,6 +229,8 @@ > > local_irq_restore(flags); > } > + > +EXPORT_SYMBOL(do_softirq); > #endif > > /* > > Patching kernel source isn't the correct fix, but bad breath is better > than no breath. >