On Thu, Jan 06, 2011 at 08:51:07PM +0000, Russell King - ARM Linux wrote:
> On Thu, Jan 06, 2011 at 12:34:32PM -0800, Tony Lindgren wrote:
> > * Tony Lindgren <[email protected]> [110106 11:52]:
> > > --- a/arch/arm/boot/compressed/head.S
> > > +++ b/arch/arm/boot/compressed/head.S
> > > @@ -71,6 +71,23 @@ wait: mrc p14, 0, pc, c0, c1, 0
> > > mov \rb, #0x50000000
> > > add \rb, \rb, #0x4000 * CONFIG_S3C_LOWLEVEL_UART_PORT
> > > .endm
> > > +#elif defined(CONFIG_ARCH_OMAP2PLUS)
> > > +#include <plat/multi.h>
> > > +#ifdef MULTI_OMAP2)
> > ^
> > Looks like my last change to this patch from if defined to ifdef
> > broke the warning above with an unbalanced bracket..
> >
> > Thanks Nishant for catching that, updated patch below.
>
> Just tried that patch, but I now have bigger problems:
>
> OMAP44XX SDP # mmcinit 0
> OMAP44XX SDP # fatload mmc 0 0x80300000 uImage-test
>
> 0 bytes read
> OMAP44XX SDP #
>
> God knows why it's doing this now, as the file is present on the SD
> card:
>
> -rwxr-xr-x. 1 rmk rmk 1836636 Jan 6 20:35 uImage-test
>
> is what's on the SD card - with no FAT errors:
>
> $ fsck.msdos /dev/mmcblk0p1
> dosfsck 3.0.1, 23 Nov 2008, FAT32, LFN
> /dev/mmcblk0p1: 8 files, 10249/142266 clusters
>
> The file's not corrupt either:
>
> $ md5sum /media/boot/uImage-test
> be3edd928f1dbb3c15215b1a8a62fa1f /media/boot/uImage-test
> $ md5sum ../build/omap4/arch/arm/boot/uImage
> be3edd928f1dbb3c15215b1a8a62fa1f ../build/omap4/arch/arm/boot/uImage
>
> Nope, still won't work:
>
> OMAP44XX SDP # setenv bootargs 'root=/dev/mmcblk0p2 rw mem=512M vmalloc=1G
> console=ttyO2,115200n8 rootdelay=2'
> OMAP44XX SDP # mmcinit 0
> OMAP44XX SDP # fatload mmc 0 0x80300000 uImage-test
>
> 0 bytes read
> OMAP44XX SDP #
>
> However, it can still boot the uImage contained on the SD card, so the
> card must be okay, and the SDP must still be able to successfully read
> from the card. I can only assume that uboot's FAT support is buggered
> in some way.
>
> We _really_ need a _decent_ boot loader on these boards, one which can
> do network booting _and_ can be configured to do so from bootup. SD
> cards just don't hack it for kernel development. It's a far too long-
> winded process for each cycle, and with all the additional problems
> (such as SD cards not being recognised 50% of the time, FAT filesystem
> bugs in boot loaders, etc) it's really not funny.
>
> So, I give up at this point with the OMAP4430SDP board. It's not worth
> spending the time with this kind of unreliability.
OMAP44XX SDP # fatls mmc 0
1935000 uimage
149104 u-boot.bin
18984 mlo
1836636 uimage-test
1130288 uimage-ss
19040 mlo.old
Invalid FAT entry
6 file(s), 0 dir(s)
"Invalid FAT entry" ? I don't think so. The thing actually contains:
-rwxr-xr-x. 1 rmk rmk 1935000 Feb 4 2010 uImage
-rwxr-xr-x. 1 rmk rmk 149104 Feb 4 2010 u-boot.bin
-rwxr-xr-x. 1 rmk rmk 18984 Jan 1 2000 mlo
-rwxr-xr-x. 1 rmk rmk 1836636 Jan 6 21:35 uImage-test
-rwxr-xr-x. 1 rmk rmk 1130288 Jan 1 2000 uImage-ss
-rwxr-xr-x. 1 rmk rmk 19040 Feb 4 2010 mlo.old
-rwxr-xr-x. 1 rmk rmk 154904 Jan 1 2000 u-boot.new
The hexdump of the directory on the SD card is:
0011a000 62 6f 6f 74 20 20 20 20 20 20 20 08 00 00 87 b6 |boot .....|
0011a010 3e 3c 3e 3c 00 00 87 b6 3e 3c 00 00 00 00 00 00 |><><....><......|
0011a020 41 75 00 49 00 6d 00 61 00 67 00 0f 00 46 65 00 |Au.I.m.a.g...Fe.|
0011a030 00 00 ff ff ff ff ff ff ff ff 00 00 ff ff ff ff |................|
0011a040 55 49 4d 41 47 45 20 20 20 20 20 20 00 64 6d 65 |UIMAGE .dme|
0011a050 44 3c 26 3e 00 00 6d 65 44 3c 4a 0e 98 86 1d 00 |D<&>..meD<J.....|
0011a060 41 75 00 2d 00 62 00 6f 00 6f 00 0f 00 30 74 00 |Au.-.b.o.o...0t.|
0011a070 2e 00 62 00 69 00 6e 00 00 00 00 00 ff ff ff ff |..b.i.n.........|
0011a080 55 2d 42 4f 4f 54 20 20 42 49 4e 20 00 64 6f 65 |U-BOOT BIN .doe|
0011a090 44 3c bc 3c 00 00 6f 65 44 3c 0e 1d 70 46 02 00 |D<.<..oeD<..pF..|
0011a0a0 41 6d 00 6c 00 6f 00 00 00 ff ff 0f 00 c8 ff ff |Am.l.o..........|
0011a0b0 ff ff ff ff ff ff ff ff ff ff 00 00 ff ff ff ff |................|
0011a0c0 4d 4c 4f 20 20 20 20 20 20 20 20 20 00 00 28 00 |MLO ..(.|
0011a0d0 21 28 26 3e 00 00 28 00 21 28 21 2d 28 4a 00 00 |!(&>..(.!(!-(J..|
0011a0e0 41 75 00 49 00 6d 00 61 00 67 00 0f 00 4e 65 00 |Au.I.m.a.g...Ne.|
0011a0f0 2d 00 74 00 65 00 73 00 74 00 00 00 00 00 ff ff |-.t.e.s.t.......|
0011a100 55 49 4d 41 47 45 7e 31 20 20 20 20 00 64 7b ac |UIMAGE~1 .d{.|
0011a110 26 3e 26 3e 00 00 7b ac 26 3e ce fb 5c 06 1c 00 |&>&>..{.&>..\...|
0011a120 e5 75 00 2d 00 62 00 6f 00 6f 00 0f 00 14 74 00 |.u.-.b.o.o....t.|
0011a130 2e 00 42 00 49 00 32 00 00 00 00 00 ff ff ff ff |..B.I.2.........|
0011a140 e5 2d 42 4f 4f 54 20 20 42 49 32 20 00 64 6f 65 |.-BOOT BI2 .doe|
0011a150 44 3c bc 3c 00 00 6f 65 44 3c 0e 1d 70 46 02 00 |D<.<..oeD<..pF..|
0011a160 41 75 00 49 00 6d 00 61 00 67 00 0f 00 6e 65 00 |Au.I.m.a.g...ne.|
0011a170 2d 00 73 00 73 00 00 00 ff ff 00 00 ff ff ff ff |-.s.s...........|
0011a180 55 49 4d 41 47 45 7e 32 20 20 20 20 00 64 bb 00 |UIMAGE~2 .d..|
0011a190 21 28 26 3e 00 00 bb 00 21 28 76 2e 30 3f 11 00 |!(&>....!(v.0?..|
0011a1a0 41 6d 00 6c 00 6f 00 2e 00 6f 00 0f 00 4e 6c 00 |Am.l.o...o...Nl.|
0011a1b0 64 00 00 00 ff ff ff ff ff ff 00 00 ff ff ff ff |d...............|
0011a1c0 4d 4c 4f 20 20 20 20 20 4f 4c 44 20 00 64 73 65 |MLO OLD .dse|
0011a1d0 44 3c 6a 3c 00 00 73 65 44 3c 32 1e 60 4a 00 00 |D<j<..seD<2.`J..|
0011a1e0 41 75 00 2d 00 62 00 6f 00 6f 00 0f 00 fa 74 00 |Au.-.b.o.o....t.|
0011a1f0 2e 00 6e 00 65 00 77 00 00 00 00 00 ff ff ff ff |..n.e.w.........|
... and the directory continues in a different sector ...
00a29200 55 2d 42 4f 4f 54 20 20 4e 45 57 20 00 00 3d 00 |U-BOOT NEW ..=.|
00a29210 21 28 26 3e 00 00 3d 00 21 28 47 2d 18 5d 02 00 |!(&>..=.!(G-.]..|
00a29220 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 |................|
The relevant parts of the FAT table are:
00004000 f8 ff ff 0f ff ff ff 0f 7b 48 00 00 00 00 00 00 |........{H......|
^^^^^ entry linking to next root-dir sector
...
000161e0 00 00 00 00 00 00 00 00 00 00 00 00 ff ff ff 0f |................|
^^^^^^^^^^^ terminator
Cluster 2 = root dir start.
Cluster 2 image address = 0x119c00 + 2 * 0x200 = 0x11a000.
Cluster 2 FAT address = 0x4000 + 2 * 4 = 0x4008.
Following cluster = 0x487b.
Cluster 0x487b image address = 0x119c00 + 0x487b * 0x200 = 0xa29200
Cluster 0x487b FAT address = 0x4000 + 0x487b * 4 = 0x161ec.
Following cluster = 0x0fffffff = FAT-32 EOF.
So, uboot's FAT code can't deal with a directory larger than one sector
that doesn't continue. While we're here, lets confirm by hand that
there's nothing wrong with the uImage-test file and it's yet again uboot
being idiotic.
The directory entry for loading my uImage-test file is:
0011a0e0 41 75 00 49 00 6d 00 61 00 67 00 0f 00 4e 65 00 |Au.I.m.a.g...Ne.|
0011a0f0 2d 00 74 00 65 00 73 00 74 00 00 00 00 00 ff ff |-.t.e.s.t.......|
0011a100 55 49 4d 41 47 45 7e 31 20 20 20 20 00 64 7b ac |UIMAGE~1 .d{.|
0011a110 26 3e 26 3e 00 00 7b ac 26 3e ce fb 5c 06 1c 00 |&>&>..{.&>..\...|
Decoding this entry gives:
creation time = 0xac7b
creation date = 0x3e26
access date = 0x3e26
time = 0xac7b
date = 0x3e26
start cluster = 0x0000fbce
size = 0x001c065c (1836636 bytes)
Cluster 0xfbce = image address 0x119c00 + 0xfbce*0x200 = 0x2093800:
Cluster 0xfbce FAT entry address = 0x4000 + 0xfbce * 4 = 0x42f38
00042f30 00 00 00 00 00 00 00 00 cf fb 00 00 d0 fb 00 00 |................|
00042f40 d1 fb 00 00 d2 fb 00 00 d3 fb 00 00 d4 fb 00 00 |................|
Next cluster = 0xfbcf, then 0xfbd0 and so on.
The data on the SD card for this file:
02093800 27 05 19 56 76 ed 63 cb 4d 26 27 76 00 1c 06 1c |'..Vv.c.M&'v....|
02093810 80 00 80 00 80 00 80 00 ad 09 bd ce 05 02 02 00 |................|
02093820 4c 69 6e 75 78 2d 32 2e 36 2e 33 37 2b 00 00 00 |Linux-2.6.37+...|
02093830 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 |................|
02093840 00 00 a0 e1 00 00 a0 e1 00 00 a0 e1 00 00 a0 e1 |................|
*
02093860 02 00 00 ea 18 28 6f 01 00 00 00 00 1c 06 1c 00 |.....(o.........|
02093870 01 70 a0 e1 02 80 a0 e1 00 20 0f e1 03 00 12 e3 |.p....... ......|
02093880 01 00 00 1a 17 00 a0 e3 56 34 12 ef 00 20 0f e1 |........V4... ..|
and the corresponding version from my build tree:
00000000 27 05 19 56 76 ed 63 cb 4d 26 27 76 00 1c 06 1c |'..Vv.c.M&'v....|
00000010 80 00 80 00 80 00 80 00 ad 09 bd ce 05 02 02 00 |................|
00000020 4c 69 6e 75 78 2d 32 2e 36 2e 33 37 2b 00 00 00 |Linux-2.6.37+...|
00000030 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 |................|
00000040 00 00 a0 e1 00 00 a0 e1 00 00 a0 e1 00 00 a0 e1 |................|
*
00000060 02 00 00 ea 18 28 6f 01 00 00 00 00 1c 06 1c 00 |.....(o.........|
00000070 01 70 a0 e1 02 80 a0 e1 00 20 0f e1 03 00 12 e3 |.p....... ......|
00000080 01 00 00 1a 17 00 a0 e3 56 34 12 ef 00 20 0f e1 |........V4... ..|
matches. So there's no problem here with the filesystem itself - it's
just uboot not parsing the FAT filesystem correctly. I *really* want
to throw this version of uboot in the circular filing cabinet now.
Can we _please_ have a version of uboot for the SDP4430 platform which
can be configured at runtime _and_ which is capable of DHCP/TFTP ?
I really don't want to mess about anymore with buggy boot loaders.
--
To unsubscribe from this list: send the line "unsubscribe linux-omap" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
