kernel_optimize_test/arch/um/drivers/chan_user.c
Jeff Dike 3d564047a5 uml: start fixing os_read_file and os_write_file
This patch starts the removal of a very old, very broken piece of code.  This
stems from the problem of passing a userspace buffer into read() or write() on
the host.  If that buffer had not yet been faulted in, read and write will
return -EFAULT.

To avoid this problem, the solution was to fault the buffer in before the
system call by touching the pages that hold the buffer by doing a copy-user of
a byte to each page.  This is obviously bogus, but it does usually work, in tt
mode, since the kernel and process are in the same address space and userspace
addresses can be accessed directly in the kernel.

In skas mode, where the kernel and process are in separate address spaces, it
is completely bogus because the userspace address, which is invalid in the
kernel, is passed into the system call instead of the corresponding physical
address, which would be valid.  Here, it appears that this code, on every host
read() or write(), tries to fault in a random process page.  This doesn't seem
to cause any correctness problems, but there is a performance impact.  This
patch, and the ones following, result in a 10-15% performance gain on a kernel
build.

This code can't be immediately tossed out because when it is, you can't log
in.  Apparently, there is some code in the console driver which depends on
this somehow.

However, we can start removing it by switching the code which does I/O using
kernel addresses to using plain read() and write().  This patch introduces
os_read_file_k and os_write_file_k for use with kernel buffers and converts
all call locations which use obvious kernel buffers to use them.  These
include I/O using buffers which are local variables which are on the stack or
kmalloc-ed.  Later patches will handle the less obvious cases, followed by a
mass conversion back to the original interface.

Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-07 12:13:03 -07:00

217 lines
5.7 KiB
C

/*
* Copyright (C) 2000 - 2003 Jeff Dike (jdike@addtoit.com)
* Licensed under the GPL
*/
#include <unistd.h>
#include <stdlib.h>
#include <errno.h>
#include <termios.h>
#include <string.h>
#include <signal.h>
#include <sched.h>
#include <sys/stat.h>
#include <sys/ioctl.h>
#include <sys/socket.h>
#include "kern_util.h"
#include "chan_user.h"
#include "user.h"
#include "os.h"
#include "choose-mode.h"
#include "mode.h"
int generic_console_write(int fd, const char *buf, int n)
{
struct termios save, new;
int err;
if(isatty(fd)){
CATCH_EINTR(err = tcgetattr(fd, &save));
if (err)
goto error;
new = save;
/* The terminal becomes a bit less raw, to handle \n also as
* "Carriage Return", not only as "New Line". Otherwise, the new
* line won't start at the first column.*/
new.c_oflag |= OPOST;
CATCH_EINTR(err = tcsetattr(fd, TCSAFLUSH, &new));
if (err)
goto error;
}
err = generic_write(fd, buf, n, NULL);
/* Restore raw mode, in any case; we *must* ignore any error apart
* EINTR, except for debug.*/
if(isatty(fd))
CATCH_EINTR(tcsetattr(fd, TCSAFLUSH, &save));
return(err);
error:
return(-errno);
}
/*
* UML SIGWINCH handling
*
* The point of this is to handle SIGWINCH on consoles which have host ttys and
* relay them inside UML to whatever might be running on the console and cares
* about the window size (since SIGWINCH notifies about terminal size changes).
*
* So, we have a separate thread for each host tty attached to a UML device
* (side-issue - I'm annoyed that one thread can't have multiple controlling
* ttys for purposed of handling SIGWINCH, but I imagine there are other reasons
* that doesn't make any sense).
*
* SIGWINCH can't be received synchronously, so you have to set up to receive it
* as a signal. That being the case, if you are going to wait for it, it is
* convenient to sit in sigsuspend() and wait for the signal to bounce you out of
* it (see below for how we make sure to exit only on SIGWINCH).
*/
static void winch_handler(int sig)
{
}
struct winch_data {
int pty_fd;
int pipe_fd;
};
static int winch_thread(void *arg)
{
struct winch_data *data = arg;
sigset_t sigs;
int pty_fd, pipe_fd;
int count, err;
char c = 1;
pty_fd = data->pty_fd;
pipe_fd = data->pipe_fd;
count = os_write_file_k(pipe_fd, &c, sizeof(c));
if(count != sizeof(c))
printk("winch_thread : failed to write synchronization "
"byte, err = %d\n", -count);
/* We are not using SIG_IGN on purpose, so don't fix it as I thought to
* do! If using SIG_IGN, the sigsuspend() call below would not stop on
* SIGWINCH. */
signal(SIGWINCH, winch_handler);
sigfillset(&sigs);
/* Block all signals possible. */
if(sigprocmask(SIG_SETMASK, &sigs, NULL) < 0){
printk("winch_thread : sigprocmask failed, errno = %d\n",
errno);
exit(1);
}
/* In sigsuspend(), block anything else than SIGWINCH. */
sigdelset(&sigs, SIGWINCH);
if(setsid() < 0){
printk("winch_thread : setsid failed, errno = %d\n", errno);
exit(1);
}
err = os_new_tty_pgrp(pty_fd, os_getpid());
if(err < 0){
printk("winch_thread : new_tty_pgrp failed, err = %d\n", -err);
exit(1);
}
/* These are synchronization calls between various UML threads on the
* host - since they are not different kernel threads, we cannot use
* kernel semaphores. We don't use SysV semaphores because they are
* persistent. */
count = os_read_file_k(pipe_fd, &c, sizeof(c));
if(count != sizeof(c))
printk("winch_thread : failed to read synchronization byte, "
"err = %d\n", -count);
while(1){
/* This will be interrupted by SIGWINCH only, since other signals
* are blocked.*/
sigsuspend(&sigs);
count = os_write_file_k(pipe_fd, &c, sizeof(c));
if(count != sizeof(c))
printk("winch_thread : write failed, err = %d\n",
-count);
}
}
static int winch_tramp(int fd, struct tty_struct *tty, int *fd_out)
{
struct winch_data data;
unsigned long stack;
int fds[2], n, err;
char c;
err = os_pipe(fds, 1, 1);
if(err < 0){
printk("winch_tramp : os_pipe failed, err = %d\n", -err);
goto out;
}
data = ((struct winch_data) { .pty_fd = fd,
.pipe_fd = fds[1] } );
/* CLONE_FILES so this thread doesn't hold open files which are open
* now, but later closed. This is a problem with /dev/net/tun.
*/
err = run_helper_thread(winch_thread, &data, CLONE_FILES, &stack, 0);
if(err < 0){
printk("fork of winch_thread failed - errno = %d\n", -err);
goto out_close;
}
*fd_out = fds[0];
n = os_read_file_k(fds[0], &c, sizeof(c));
if(n != sizeof(c)){
printk("winch_tramp : failed to read synchronization byte\n");
printk("read failed, err = %d\n", -n);
printk("fd %d will not support SIGWINCH\n", fd);
err = -EINVAL;
goto out_close;
}
return err ;
out_close:
os_close_file(fds[1]);
os_close_file(fds[0]);
out:
return err;
}
void register_winch(int fd, struct tty_struct *tty)
{
int pid, thread, thread_fd = -1;
int count;
char c = 1;
if(!isatty(fd))
return;
pid = tcgetpgrp(fd);
if(!CHOOSE_MODE_PROC(is_tracer_winch, is_skas_winch, pid, fd,
tty) && (pid == -1)){
thread = winch_tramp(fd, tty, &thread_fd);
if(thread > 0){
register_winch_irq(thread_fd, fd, thread, tty);
count = os_write_file_k(thread_fd, &c, sizeof(c));
if(count != sizeof(c))
printk("register_winch : failed to write "
"synchronization byte, err = %d\n",
-count);
}
}
}
/*
* Overrides for Emacs so that we follow Linus's tabbing style.
* Emacs will notice this stuff at the end of the file and automatically
* adjust the settings for this buffer only. This must remain at the end
* of the file.
* ---------------------------------------------------------------------------
* Local variables:
* c-file-style: "linux"
* End:
*/