On Sun, Dec 27, 2015 at 10:45:18AM -0700, Roy Bixler wrote:
> On Sat, Dec 26, 2015 at 10:54:34PM +0900, Makoto Fujiwara wrote:
> > options MSGBUFSIZE=1048576
>
> Good idea. I put this in my NOUVEAU kernel configuration and was able
> to get some more dmesg output. I'm not set up with a serial console
> and there is voluminous output on bootup, so I'll just attach a few
> screenshots. The first one shows that the nouveau0 device attaches to
> the card and doesn't show any errors. The second one shows that the
> initialisation of the nouveau driver appears to be successful and
> finds its PRAMIN image. The final image shows the failure to "map
> PRAMIN BAR". I put an additional debug message in the code in
> /usr/src/sys/external/bsd/drm2/dist/drm/nouveau/core/subdev/instmem/nouveau_subdev_instmem_nv40.c:
>
> #ifdef __NetBSD__
> priv->iomemt = nv_device_resource_tag(device, bar);
> priv->iomemsz = nv_device_resource_len(device, bar);
> nv_error(priv, "bar = %d, iomemt=%x, iomemsz=%x,
> start=%x\n", bar, (unsigned int) priv->iomemt, (unsigned int)
> priv->iomemsz, (unsigned int) nv_device_resource_start(device,
> bar));
> if (bus_space_map(priv->iomemt,
> nv_device_resource_start(device, bar),
> priv->iomemsz, 0, &priv->iomemh)) {
> priv->iomemsz = 0;
> nv_error(priv, "unable to map PRAMIN BAR\n");
> return -EFAULT;
> }
>
> which shows that
>
> bar = 2, iomemt=c1077780, iomemsz=10, start=0
I tried a further experiment with some intriguing results. With the
attached patch, which sets "bar = 1" and adds debugging statements
with some delays, I at least got the "nv40_instmem_ctor()" function to
succeed. The result of the above statement also looks more sensible:
bar = 1, iomemt=c107c700, iomemsz=1000000, start=c0000000
Afterwards, the video mode changes, but then I only get a blank
screen. Screen shots showing the debug output also attached. I
suppose that, to make any further progress, I'll need to get a null
modem cable and set up a serial console.
--
Roy Bixler <rcbixler%nyx.net@localhost>
"The fundamental principle of science, the definition almost, is this: the
sole test of the validity of any idea is experiment."
-- Richard P. Feynman
--- nouveau_subdev_instmem_nv40.c.orig 2015-11-12 11:34:08.000000000 +0000
+++ nouveau_subdev_instmem_nv40.c 2015-12-29 15:21:39.000000000 +0000
@@ -76,10 +76,13 @@
bar = 2;
else
bar = 3;
+ bar = 1;
#ifdef __NetBSD__
priv->iomemt = nv_device_resource_tag(device, bar);
priv->iomemsz = nv_device_resource_len(device, bar);
+nv_error(priv, "bar = %d, iomemt=%x, iomemsz=%x, start=%x\n", bar, (unsigned int) priv->iomemt, (unsigned int) priv->iomemsz, (unsigned int) nv_device_resource_start(device, bar));
+mdelay(5000);
if (bus_space_map(priv->iomemt, nv_device_resource_start(device, bar),
priv->iomemsz, 0, &priv->iomemh)) {
priv->iomemsz = 0;
@@ -94,6 +97,8 @@
return -EFAULT;
}
#endif
+nv_error(priv, "ioremap succeeded\n");
+mdelay(1000);
/* PRAMIN aperture maps over the end of vram, reserve enough space
* to fit graphics contexts for every channel, the magics come
@@ -115,18 +120,24 @@
if (ret)
return ret;
+nv_error(priv, "mm_init succeeded\n");
+mdelay(1000);
/* 0x00000-0x10000: reserve for probable vbios image */
ret = nouveau_gpuobj_new(nv_object(priv), NULL, 0x10000, 0, 0,
&priv->vbios);
if (ret)
return ret;
+nv_error(priv, "gpuobj succeeded\n");
+mdelay(1000);
/* 0x10000-0x18000: reserve for RAMHT */
ret = nouveau_ramht_new(nv_object(priv), NULL, 0x08000, 0,
&priv->ramht);
if (ret)
return ret;
+nv_error(priv, "ramht succeeded\n");
+mdelay(15000);
/* 0x18000-0x18200: reserve for RAMRO
* 0x18200-0x20000: padding
*/
@@ -135,6 +146,8 @@
if (ret)
return ret;
+nv_error(priv, "gpuobj2 succeeded\n");
+mdelay(1000);
/* 0x20000-0x21000: reserve for RAMFC
* 0x21000-0x40000: padding and some unknown crap
*/
@@ -142,6 +155,8 @@
NVOBJ_FLAG_ZERO_ALLOC, &priv->ramfc);
if (ret)
return ret;
+nv_error(priv, "instmem_ctor succeeded\n");
+mdelay(15000);
return 0;
}
Attachment:
20151229_092310.jpeg
Description: JPEG image
Attachment:
20151229_091846.jpeg
Description: JPEG image
