Subject: Re: Understanding Alpha syscalls
From: B. James Phillippe (bryan@terran.org)
Date: Wed Jun 02 1999 - 19:15:39 PDT
On Wed, 2 Jun 1999, Uncle George wrote:
> u mean u cant "find /usr/src/linux -type f -exec grep -l sys_open {} \;"
> for sys_open in fs/open.c ?
> gat
I did find it there, but I didn't understand the mapping between the
assembly and the C... maybe you can help me. Confirm if I understand it
correctly now:
1.) "open" called from user-space program.
2.) "open" defined in some way within libc such that it uses the syscall2
(or 3) macro defined by <asm/unistd.h>. ?
3.) syscallx expands to an assembly function which calls the PALcode with
"callsys" with the appropriate __NR_sycallname and arguments (I am an
assembly newbie and I know nothing of PAL).
4.) kernel has registered (at bootup, in trap_init()) that the routine
"entSys" is to be called by the PALcode for system calls. ?
5.) PALcode traps into kernel to execute the asm routine entSys. ?
6.) entSys runs the function in the sys_call_table at the offset X, where X
was the name of the __NR_syscallname (which is actuallly a number). ?
7.) the function in the sys_call_table can be a straight C-code routine,
such as the one you pointed out is in fs/open.c for sys_open.
8.) when the function returns, the stack is unwound and the rest of the
entSys stuff is done, including passing the return value back up to
user-space. ?
Is this right?
thanks for the help,
-bp
-- # Software Development, WatchGuard Inc. # Projects, PGP key at http://www.terran.org/~bryan-- To unsubscribe: send e-mail to axp-list-request@redhat.com with 'unsubscribe' as the subject. Do not send it to axp-list@redhat.com
This archive was generated by hypermail 2a22 : Wed Jun 30 1999 - 19:27:25 PDT