Skip to contentSkip to conversation

Little bug in O(1) scheduler patch (also in 2.4.19-ac4)

DISCUSSION Juan M. de la Torre 2 participants 3 messages Sep 6, 2002

lore
Share & tools
lore
#0 lore raw

Hi,

a module compiled for kernel 2.4.19-ac4, with CONFIG_MODVERSIONS, and
importing flush_signals() from kernel, fails to load, reporting "unresolved symbol flush_signals_xxxxxxxx".

The problem is that the type of the argument passed to flush_signals()
has been changed from "struct task_struct *" to "task_t *" in sched.h, but it remains unchanged in kernel/signal.c. The same happens with flush_signal_handlers().

Sorry if this is a known issue.

Regargs,
Juanma

PS: First sent to Ingo Molnar, since no response, reposting here again.

Signature · 1 line
/jm
#1 lore raw
Quoted 3 lines · Juan M. de la Torre

a module compiled for kernel 2.4.19-ac4, with CONFIG_MODVERSIONS, and
importing flush_signals() from kernel, fails to load, reporting "unresolved symbol flush_signals_xxxxxxxx".

what module?

Quoted 4 lines

The problem is that the type of the argument passed to flush_signals()
has been changed from "struct task_struct *" to "task_t *" in sched.h, but it remains unchanged in kernel/signal.c. The same happens with flush_signal_handlers().

hm, struct task_struct * is the same as task_t *.

Ingo

#2 lore raw

I was wrong; the problem i was experimenting was due to a symbolic link
problem (i was compiling the module with other kernel headers). Since the CRC of a versioned symbol is based on the function arguments i thought that the different type of the argument (well, is the same type with different name) has something to do with the error i was getting.

I apologize for wasting readers' time.

Regards,
Juanma

Signature · 1 line
/jm