[patch] swarm-cs4297a: Support little-endian configuration

[patch] swarm-cs4297a: Support little-endian configuration

[Maciej W. Rozycki]

Hello,

 The swarm-cs4297a driver currently works only with a big-endian
configuration -- if run in little-endian one, it floods the console with
an infinite stream of error messages, making the kernel effectively
unusable.

 Here's a fix.  Apart from removing the flood, it actually makes audio
work, at least the output.  Fixes for the input are untested, due to a
lack of a suitable input device, but hopefully they are OK.

 There's one functional difference -- the original code tried to
sign-extend 16-bit samples to 20 bits.  It did it incorrectly and also the
AC'97 spec says to send bits missing from samples as zeros, so I've 
removed that fragment.

 OK to apply?

  Maciej

-- 
+  Maciej W. Rozycki, Technical University of Gdansk, Poland   +
+--------------------------------------------------------------+
+        e-mail: macro@ds2.pg.gda.pl, PGP key available        +

patch-mips-2.4.24-pre2-20040116-swarm-cs4297a-3
diff -up --recursive --new-file linux-mips-2.4.24-pre2-20040116.macro/drivers/sound/swarm_cs4297a.c linux-mips-2.4.24-pre2-20040116/drivers/sound/swarm_cs4297a.c
--- linux-mips-2.4.24-pre2-20040116.macro/drivers/sound/swarm_cs4297a.c	2003-07-03 02:57:00.000000000 +0000
+++ linux-mips-2.4.24-pre2-20040116/drivers/sound/swarm_cs4297a.c	2004-04-05 05:48:34.000000000 +0000
@@ -11,6 +11,7 @@
 *            -- adapted from cs4281 PCI driver for cs4297a on
 *               BCM1250 Synchronous Serial interface
 *               (Kip Walker, Broadcom Corp.)
+*      Copyright (C) 2004  Maciej W. Rozycki
 *
 *      This program is free software; you can redistribute it and/or modify
 *      it under the terms of the GNU General Public License as published by
@@ -71,14 +72,16 @@
 #include <linux/ac97_codec.h>
 #include <linux/pci.h>
 #include <linux/bitops.h>
-#include <asm/io.h>
-#include <asm/dma.h>
 #include <linux/init.h>
 #include <linux/poll.h>
 #include <linux/smp_lock.h>
 #include <linux/wrapper.h>
-#include <asm/uaccess.h>
+
+#include <asm/byteorder.h>
+#include <asm/dma.h>
 #include <asm/hardirq.h>
+#include <asm/io.h>
+#include <asm/uaccess.h>
 
 #include <asm/sibyte/sb1250_regs.h>
 #include <asm/sibyte/sb1250_int.h>
@@ -758,7 +761,7 @@ static int serdma_reg_access(struct cs42
 
                 descr = &d->descrtab[swptr];
                 data_p = &d->dma_buf[swptr * 4];
-                *data_p = data;
+		*data_p = cpu_to_be64(data);
                 out64(1, SS_CSR(R_SER_DMA_DSCR_COUNT_TX));
                 CS_DBGOUT(CS_DESCR, 4,
                           printk(KERN_INFO "cs4297a: add_tx  %p (%x -> %x)\n",
@@ -950,7 +953,7 @@ static void cs4297a_update_ptr(struct cs
                         s_ptr = (u32 *)&(d->dma_buf[d->swptr*4]);
                         descr = &d->descrtab[d->swptr];
                         while (diff2--) {
-                                u64 data = *(u64 *)s_ptr;
+				u64 data = be64_to_cpu(*(u64 *)s_ptr);
                                 u64 descr_a;
                                 u16 left, right;
                                 descr_a = descr->descr_a;
@@ -977,10 +980,11 @@ static void cs4297a_update_ptr(struct cs
                                         continue;
                                 }
                                 good_diff++;
-                                left = ((s_ptr[1] & 0xff) << 8) | ((s_ptr[2] >> 24) & 0xff);
-                                right = (s_ptr[2] >> 4) & 0xffff;
-                                *d->sb_hwptr++ = left;
-                                *d->sb_hwptr++ = right;
+				left = ((be32_to_cpu(s_ptr[1]) & 0xff) << 8) |
+				       ((be32_to_cpu(s_ptr[2]) >> 24) & 0xff);
+				right = (be32_to_cpu(s_ptr[2]) >> 4) & 0xffff;
+				*d->sb_hwptr++ = cpu_to_be16(left);
+				*d->sb_hwptr++ = cpu_to_be16(right);
                                 if (d->sb_hwptr == d->sb_end)
                                         d->sb_hwptr = d->sample_buf;
                                 descr++;
@@ -1025,7 +1029,7 @@ static void cs4297a_update_ptr(struct cs
                            here because of an interrupt, so there must
                            be a buffer to process. */
                         do {
-                                data = *data_p;
+				data = be64_to_cpu(*data_p);
                                 if ((descr->descr_a & M_DMA_DSCRA_A_ADDR) != PHYSADDR((long)data_p)) {
                                         printk(KERN_ERR "cs4297a: RX Bad address %d (%llx %lx)\n", d->swptr,
                                                (long long)(descr->descr_a & M_DMA_DSCRA_A_ADDR),
@@ -1804,7 +1808,6 @@ static ssize_t cs4297a_write(struct file
                 u32 *s_tmpl;
                 u32 *t_tmpl;
                 u32 left, right;
-                /* XXXKW check system endian here ... */
                 int swap = (s->prop_dac.fmt == AFMT_S16_LE) || (s->prop_dac.fmt == AFMT_U16_LE);
                 
                 /* XXXXXX this is broken for BLOAT_FACTOR */
@@ -1845,21 +1848,21 @@ static ssize_t cs4297a_write(struct file
 
                 /* XXXKW assuming 16-bit stereo! */
                 do {
-                        t_tmpl[0] = 0x98000000;
-                        left = s_tmpl[0] >> 16;
-                        if (left & 0x8000)
-                                left |= 0xf0000;
-                        right = s_tmpl[0] & 0xffff;
-                        if (right & 0x8000)
-                                right |= 0xf0000;
-                        if (swap) {
-                          t_tmpl[1] = left & 0xff;
-                          t_tmpl[2] = ((left & 0xff00) << 16) | ((right & 0xff) << 12) |
-                              ((right & 0xff00) >> 4);
-                        } else {
-                          t_tmpl[1] = left >> 8;
-                          t_tmpl[2] = ((left & 0xff) << 24) | (right << 4);
-                        }
+			u32 tmp;
+
+			t_tmpl[0] = cpu_to_be32(0x98000000);
+
+			tmp = be32_to_cpu(s_tmpl[0]);
+			left = tmp & 0xffff;
+			right = tmp >> 16;
+			if (swap) {
+				left = swab16(left);
+				right = swab16(right);
+			}
+			t_tmpl[1] = cpu_to_be32(left >> 8);
+			t_tmpl[2] = cpu_to_be32(((left & 0xff) << 24) |
+						(right << 4));
+
                         s_tmpl++;
                         t_tmpl += 8;
                         copy_cnt -= 4;
@@ -1867,7 +1870,8 @@ static ssize_t cs4297a_write(struct file
 
                 /* Mux in any pending read/write accesses */
                 if (s->reg_request) {
-                        *(u64 *)(d->dma_buf + (swptr * 4)) |= s->reg_request;
+			*(u64 *)(d->dma_buf + (swptr * 4)) |=
+				cpu_to_be64(s->reg_request);
                         s->reg_request = 0;
                         wake_up(&s->dma_dac.reg_wait);
                 }

Re: [patch] swarm-cs4297a: Support little-endian configuration

[Martin Michlmayr]

* Maciej W. Rozycki <macro@ds2.pg.gda.pl> [2004-04-05 13:09]:
> The swarm-cs4297a driver currently works only with a big-endian
> configuration -- if run in little-endian one, it floods the console with
> an infinite stream of error messages, making the kernel effectively

Since you seem to use a SWARM in little-endian, let me ask 2 questions.
Are you using their boot loader (sibyl), and did you notice it has
problems in little-endian?  (IF so, do you have patches?  ;-)  Also,
have you been able to compile sibyl against a normal e2fslibs rather
than their custom version?
-- 
Martin Michlmayr
tbm@cyrius.com

Re: [patch] swarm-cs4297a: Support little-endian configuration

[Maciej W. Rozycki]

On Mon, 5 Apr 2004, Martin Michlmayr wrote:

> Are you using their boot loader (sibyl), and did you notice it has
> problems in little-endian?  (IF so, do you have patches?  ;-)  Also,
> have you been able to compile sibyl against a normal e2fslibs rather
> than their custom version?

 I boot kernels directly (a bit dissatisfied with no ELF64 support in CFE,
but that can be fixed, I suppose) over the network, so I can't help you
with your problems, sorry.

-- 
+  Maciej W. Rozycki, Technical University of Gdansk, Poland   +
+--------------------------------------------------------------+
+        e-mail: macro@ds2.pg.gda.pl, PGP key available        +

Re: [patch] swarm-cs4297a: Support little-endian configuration

[Jun Sun]

On Mon, Apr 05, 2004 at 03:01:19PM +0200, Maciej W. Rozycki wrote:
> On Mon, 5 Apr 2004, Martin Michlmayr wrote:
> 
> > Are you using their boot loader (sibyl), and did you notice it has
> > problems in little-endian?  (IF so, do you have patches?  ;-)  Also,
> > have you been able to compile sibyl against a normal e2fslibs rather
> > than their custom version?
> 
>  I boot kernels directly (a bit dissatisfied with no ELF64 support in CFE,
> but that can be fixed, I suppose) over the network, so I can't help you
> with your problems, sorry.
>

I have been using objcopy to convert ELF64 to ELF32 and then boot through 
CFE (suggested by Drow).  This seems to be working fine.

I have also used a LE version sibyl before and did not notice any problem.
I think that version is a binary from broadcom.

Jun

Re: [patch] swarm-cs4297a: Support little-endian configuration

[Maciej W. Rozycki]

On Mon, 5 Apr 2004, Jun Sun wrote:

> I have been using objcopy to convert ELF64 to ELF32 and then boot through 
> CFE (suggested by Drow).  This seems to be working fine.

 It does work and the Linux Makefile even does the conversion
automatically.  It's a bit annoying, though, to have to keep two images.

-- 
+  Maciej W. Rozycki, Technical University of Gdansk, Poland   +
+--------------------------------------------------------------+
+        e-mail: macro@ds2.pg.gda.pl, PGP key available        +

Re: [patch] swarm-cs4297a: Support little-endian configuration

[Mitch Lichtenberg]

Adding ELF64 to CFE has been on my to-do list for ages.   Thanks for
the reminder, I'll try to bump it up a few notches.
Now I know who to send it to for testing :-).

/Mitch.


> On Mon, 5 Apr 2004, Jun Sun wrote:
> 
> 
>>I have been using objcopy to convert ELF64 to ELF32 and then boot through 
>>CFE (suggested by Drow).  This seems to be working fine.
> 
> 
>  It does work and the Linux Makefile even does the conversion
> automatically.  It's a bit annoying, though, to have to keep two images.
> 

Re: [patch] swarm-cs4297a: Support little-endian configuration

[Maciej W. Rozycki]

On Mon, 5 Apr 2004, Mitch Lichtenberg wrote:

> Adding ELF64 to CFE has been on my to-do list for ages.   Thanks for
> the reminder, I'll try to bump it up a few notches.

 Note it may be especially useful if booting a kernel (or any other
binary) to be loaded into XKPHYS.

-- 
+  Maciej W. Rozycki, Technical University of Gdansk, Poland   +
+--------------------------------------------------------------+
+        e-mail: macro@ds2.pg.gda.pl, PGP key available        +