Hi, My aim was to place some code inside entry.S such that when a user space application invokes a system call I can do some instrumentation related work like taking the reading of the 64-bit time stamp etc. etc. My code can be found at the end of this email. Now my code seems to work fine on a uni processor kernel, but gives me some funny errors on a SMP kernel. My code gets activated when I write 1 to /proc/sched/start. At that point hn_trace_start gets toggled to 1. echo "1" > /proc/sched/start As soon as I do this, I get the following weird errors when I invoke commands which internally call a system call (ls, rm, cat, anything!). <errors> [root@llm05 /]# rm -bash: child setpgid (-38 to -38): Invalid argument [root@llm05 root]# ls -bash: child setpgid (-38 to -38): Invalid argument ls: write error: Function not implemented [root@llm05 hareesh]# cat /proc/sched/cpu0 -bash: child setpgid (-38 to -38): Invalid argument cat: /proc/sched/cpu0: Function not implemented [root@llm05 hareesh]# reboot -bash: child setpgid (-38 to -38): Invalid argument </errors> What could be the problem? Should I place my code at a 'safer' point inside entry.S? Thank you. -- Hareesh Nagarajan WEB: http://puggy.symonds.net/~hareesh <code> cmpl $__PAGE_OFFSET-3,%ebp jae syscall_fault 1: movl (%ebp),%ebp .section __ex_table,"a" .align 4 .long 1b,syscall_fault .previous pushl %eax SAVE_ALL GET_THREAD_INFO(%ebp) /*-------------- MY CODE BEGINS ------------------*/ cmpl $0, hn_trace_start # Allow tracing? je .no_start GET_CPU(%ecx) GET_RECORD_SIZE(%ecx) GET_RECORD($hn_arr, %ecx, %ebx) addl $BUF_SIZE, %ebx movl $0, %ecx cmpl %ecx, (%ebx, %ecx, 4) jne .no_start addl $4, %ebx cmpl %ecx, (%ebx, %ecx, 4) jne .no_start rdtsc movl %edx, (%ebx, %ecx, 4) subl $4, %ebx movl %eax, (%ebx, %ecx, 4) .no_start: /*-------------- MY CODE ENDS ------------------*/ </code> -- Kernelnewbies: Help each other learn about the Linux kernel. Archive: http://mail.nl.linux.org/kernelnewbies/ FAQ: http://kernelnewbies.org/faq/
