NCR53c406a: Apparently wait_intr() is unused, so remove it.
--- linux-2.6.0-test5/drivers/scsi/NCR53c406a.c.orig Sun Aug 24 09:49:34 2003
+++ linux-2.6.0-test5/drivers/scsi/NCR53c406a.c Tue Sep 9 15:01:48 2003
@@ -170,7 +170,6 @@
/* Static function prototypes */
static void NCR53c406a_intr(int, void *, struct pt_regs *);
static irqreturn_t do_NCR53c406a_intr(int, void *, struct pt_regs *);
-static void wait_intr(void);
static void chip_init(void);
static void calc_port_addr(void);
#ifndef IRQ_LEV
@@ -663,26 +662,6 @@
{
DEB(printk("NCR53c406a_info called\n"));
return (info_msg);
-}
-
-static void wait_intr(void)
-{
- unsigned long i = jiffies + WATCHDOG;
-
- while (time_after(i, jiffies) && !(inb(STAT_REG) & 0xe0)) { /* wait for a pseudo-interrupt */
- cpu_relax();
- barrier();
- }
-
- if (time_before_eq(i, jiffies)) { /* Timed out */
- rtrc(0);
- current_SC->result = DID_TIME_OUT << 16;
- current_SC->SCp.phase = idle;
- current_SC->scsi_done(current_SC);
- return;
- }
-
- NCR53c406a_intr(0, NULL, NULL);
}
static int NCR53c406a_queue(Scsi_Cmnd * SCpnt, void (*done) (Scsi_Cmnd *))
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- Sony Network and Software Technology Center Europe (NSCE)
[email protected] ------------------- Sint-Stevens-Woluwestraat 55
Voice +32-2-2908453 Fax +32-2-7262686 ---------------- B-1130 Brussels, Belgium
On Thu, 2003-09-11 at 02:37, Geert Uytterhoeven wrote:
> NCR53c406a: Apparently wait_intr() is unused, so remove it.
It is currently unused. However, the reason is that we removed the scsi
command method that allows polled operation in a driver (this routine is
actually polling the interrupt port on the chip).
I'd like to wait a while to see if anyone still needs this mode when 2.6
gets a wider test audience. If you wish, you can surround the routine
with #if 0 and a comment saying we can junk it later if it really is
unnecessary.
Thanks,
James
On 12 Sep 2003, James Bottomley wrote:
> > NCR53c406a: Apparently wait_intr() is unused, so remove it.
>
> It is currently unused. However, the reason is that we removed the scsi
> command method that allows polled operation in a driver (this routine is
> actually polling the interrupt port on the chip).
>
> I'd like to wait a while to see if anyone still needs this mode when 2.6
> gets a wider test audience. If you wish, you can surround the routine
> with #if 0 and a comment saying we can junk it later if it really is
> unnecessary.
I've encountered an ISA adapter using this chip in polled mode (no ISA
IRQ line routed to the chip) quite recently. But I can't say if the guy
using it won't throw it away before final 2.6. ;-)
--
+ Maciej W. Rozycki, Technical University of Gdansk, Poland +
+--------------------------------------------------------------+
+ e-mail: [email protected], PGP key available +
On Gwe, 2003-09-12 at 15:16, Maciej W. Rozycki wrote:
> I've encountered an ISA adapter using this chip in polled mode (no ISA
> IRQ line routed to the chip) quite recently. But I can't say if the guy
> using it won't throw it away before final 2.6. ;-)
For a lot of these dumb controllers polled I/O seems to materially
outperform interrupt driven I/O on SMP boxes - its true of the NCR5380
for example. We trade a ton of locking jamming the box up for a CPU
thats spending some of its time pretending to be half a scsi chip.