RE: Proposed patch tty_io.c/8K write split

heffnerd (Fabian.Frederick@prov-liege.be)
Tue, 24 Aug 1999 18:28:37 +0200

-> I have previously posted a suggestion to allow line disciplines
-> to avoid the splitting of write() calls into 8K chunks. No
-> one objected
-> so I am posting the actual patch for comments. The patch is
-> against 2.2.11.
-> The point of this change is to allow a line discipline I wrote to
-> process complete write() calls without interference from tty_io.c.
->
-> The change adds a new TTY flag that suppresses the write split
-> in tty_io.c. If this flag is not set then the behavior is
-> unchanged. If set
-> then the line discipline write routine is called with the
-> parameters of the
-> original write (no write split). When a line discipline is
-> loaded (ldisc
-> open called)
-> then the ldisc can set the flag if necessary. When the ldisc
-> is unloaded
-> (ldisc close called), then the flag can be restored.
->
-> This patch is very minimal and does not change the default
-> behavior if the
-> new flag is not set. Please post any objections. If none are
-> raised then
-> I will submit this patch.
->
-> diff -Naur linux-2.2.11/drivers/char/tty_io.c
-> linux-2.2.11-mg/drivers/char/tty_io.c
-> --- linux-2.2.11/drivers/char/tty_io.c Mon Aug 9 14:05:09 1999
-> +++ linux-2.2.11-mg/drivers/char/tty_io.c Mon Aug 23 15:36:41 1999
-> @@ -656,23 +656,27 @@
-> down(&inode->i_sem);
-> return -ERESTARTSYS;
-> }
-> - for (;;) {
-> - unsigned long size = PAGE_SIZE*2;
-> - if (size > count)
-> - size = count;
-> - ret = write(tty, file, buf, size);
-> - if (ret <= 0)
-> - break;
-> - written += ret;
-> - buf += ret;
-> - count -= ret;
-> - if (!count)
-> - break;
-> - ret = -ERESTARTSYS;
-> - if (signal_pending(current))
-> - break;
-> - if (current->need_resched)
-> - schedule();

-> + if ( test_bit(TTY_NO_WRITE_SPLIT, &tty->flags) )

Why playing with flags here ?...
if (!count%8192)
/*process one frame UNINTERRUPTIBLE*/
-> + written = write(tty, file, buf, count);
/*Check 'written' here not below ....
that's one test non 8K condition won't race. */

-> + else {
-> + for (;;) {
-> + unsigned long size = PAGE_SIZE*2;
-> + if (size > count)
-> + size = count;
-> + ret = write(tty, file, buf, size);
-> + if (ret <= 0)
-> + break;
-> + written += ret;
-> + buf += ret;
-> + count -= ret;
-> + if (!count)
-> + break;
-> + ret = -ERESTARTSYS;
-> + if (signal_pending(current))
-> + break;
-> + if (current->need_resched)
-> + schedule();
-> + }
-> }
-> if (written) {
-> file->f_dentry->d_inode->i_mtime = CURRENT_TIME;
-> diff -Naur linux-2.2.11/include/linux/tty.h
-> linux-2.2.11-mg/include/linux/tty.h
-> --- linux-2.2.11/include/linux/tty.h Mon Aug 23 14:52:57 1999
-> +++ linux-2.2.11-mg/include/linux/tty.h Mon Aug 23 15:34:43 1999
-> @@ -328,6 +328,7 @@
-> #define TTY_HW_COOK_OUT 14
-> #define TTY_HW_COOK_IN 15
-> #define TTY_PTY_LOCK 16
-> +#define TTY_NO_WRITE_SPLIT 17
->
-> #define TTY_WRITE_FLUSH(tty) tty_write_flush((tty))
->
->
-> Paul Fulghum paulkf@microgate.com
-> Microgate Corporation www.microgate.com
-> 9501 Capital of Texas Hwy
-> Austin, Texas 78759
-> (512)-345-7791
->
->
-> -
-> To unsubscribe from this list: send the line "unsubscribe
-> linux-kernel" in
-> the body of a message to majordomo@vger.rutgers.edu
-> Please read the FAQ at http://www.tux.org/lkml/
->

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/