<lekernel>
I guess I'll need to completely rewrite that fucking driver if I want it to work. sigh.
<adamw_>
lekernel, interesting thing: Do you remember Uwe Bonnes's llbbc board? It comes with the same idea with our emails about using PROGRAM_B & FLASH_RESET_N as AND gate's inputs, then AND's output connects to flash reset pin.
<adamw_>
the AND gate is acted exactly with two equivalent diodes inside a gate. hmm..i should study more details. :-)
<Fallenou>
lekernel: yes I saw the e-mail
<Fallenou>
shame on me
<Fallenou>
I'm changing that today
<Fallenou>
lekernel:Â Â fixed pushed
<CIA-94>
rtems-milkymist: Yann Sionneau master * r459b785 / c/src/lib/libbsp/lm32/shared/milkymist_networking/network.c : fixed dma txbuffer allocated on the stack - http://bit.ly/gbfm2r
<Fallenou>
kristianpaul: can you pull from rtems-milkymist , compile and test again to see if rx fifo overflow is still there ? :)
<Fallenou>
thanks
<lekernel>
anyone has still an ISE 11.x install here?
<lekernel>
wolfspraul: are the assembled MM1s (with case) shippable now?
<wolfspraul>
ha, just about to send an email
<wolfspraul>
I will not change to shipping assembled units, I will just add the unassembled case kit into the box for now
<wolfspraul>
the reason is that we only have a handful left anyway, and most likely the people buying now won't mind having to follow the assembly instructions
<wolfspraul>
if I switch to fully assembled product, I have to make a new box and label, which I will do but there is no point in rushing it now
<kristianpaul>
Fallenou: damm not now, i mean i can but i dont have a MM1 with me at work, but i'll do it tonight
<kristianpaul>
actually i'm eading you mail in the rtmes ML now ;-)
<kristianpaul>
Reading*
<Fallenou>
ok :p
<Fallenou>
great keep me informed of how the test are going
<lekernel>
xc4010... I bet the DRAM controller would fill one
<CIA-94>
milkymist: Sebastien Bourdeauducq master * ra20c740 / software/libbase/blockdev.c : Memory card R2 response is only 17 bytes long, thanks Michael Walle for pointing that out - http://bit.ly/gIm02u
<CIA-94>
rtems-milkymist: Sebastien Bourdeauducq master * ra8a908e / c/src/lib/libbsp/lm32/shared/milkymist_memcard/memcard.c : Memory card R2 response is only 17 bytes long, thanks Michael Walle for pointing that out - http://bit.ly/hjni0x
<Fallenou>
kristianpaul: any luck with the network driver ?
<kristianpaul>
Fallenou: is GMT -5 here, sorry
<kristianpaul>
i still at work..
<kristianpaul>
80Km from home
<kristianpaul>
by then milkymist
<Fallenou>
oh ok :p
<kristianpaul>
Fallenou: yeah i feel sorry our time difference is not good for this tests.
<kristianpaul>
But tomorrow i dont have to work so i can be all day at home doing my personal stuff more freely  :-)
<Fallenou>
hehe
<Fallenou>
tomorrow and this week-end I won't have much time for hacking
<kristianpaul>
:-//
<Fallenou>
my girl friend come in london to see me and visit :)
<Fallenou>
but don't feel sorry, you use your precious time to help
<kristianpaul>
Fallenou: is your fix  about  the DMA thing?
<Fallenou>
yes
<kristianpaul>
good
<kristianpaul>
I dint check the diffs.. long time, lots of mails to answer..
<kristianpaul>
ah,girl
<Fallenou>
as long as we can tests and leave messages here it's good :)
<kristianpaul>
free of compromises
<kristianpaul>
sure, as always
<kristianpaul>
I'll repeat test, publish captures and logs
<Fallenou>
great, awesome !
<Fallenou>
thanks
<Fallenou>
I had dma problems too at work today :p
<kristianpaul>
Fallenou: btw let said i have a custom milkyist core, it is already maped on memory of course, do i need a driver or i can map it directly on rtems?
<Fallenou>
I am doing some DMA with a pic32
<kristianpaul>
pic32, nice
<Fallenou>
if you just want to write/read to your registers
<Fallenou>
you can do it in an application
<Fallenou>
no need for driver
<kristianpaul>
I works with pics since school, but just up 16bits..
<Fallenou>
but writting a driver is cleaner :)
<kristianpaul>
sure :-)
<Fallenou>
but for tests, a little app is enough
<Fallenou>
it's like the bios :)
<kristianpaul>
well bios dont have yaffs supports, i need dump content somwhere..
<kristianpaul>
and read it later, no matter if ethernet dint work, or SD bug, i need read it back at the end..
<Fallenou>
sure use rtems capabilities
<mwalle>
hi
<kristianpaul>
hey mwalle
<Fallenou>
hi mwalle :)
<Fallenou>
you flood my mailbox with patches that's great mwalle :D
<kristianpaul>
hehe
<mwalle>
larsc: i've create a uclibc repo, uploaded the toolchain patches to the mm wiki and posted some patches to the ML
<mwalle>
Fallenou: :b
<lekernel>
ok, let's try bisecting that gcc 4.6 bug now...
<lekernel>
is unstacking a heap of small tasks that have been around for too long
<larsc>
mwalle: great, thanks :)
<kristianpaul>
lekernel: hey, thanks for the SD fix, i-ll try  later too :-)
<Fallenou>
ahah you have two fixes to test
<Fallenou>
happy ? :p
<mwalle>
kristianpaul: keep in mind that this probably won't fix the flooding when no sd card image is provided
<kristianpaul>
mwalle: k
<mwalle>
gn8
<Fallenou>
gn8
<Fallenou>
kristianpaul: how did you create the yaffs filesystem on the sd card ?
<Fallenou>
or is it another file system to use ?
<Fallenou>
I don't think my ubuntu has support for yaffs2
<lekernel>
it's probably another fs, it would be pretty hard to use yaffs on a sd card
<lekernel>
yaffs uses out-of-band NAND data - getting it to work with memories without OOB (like NOR) requires a bit of coding
<Fallenou>
oh ok I mixed up
<Fallenou>
what file system do I have to use on the file I give to qemu -sdcard ?
<lekernel>
FAT
<Fallenou>
fat32 ?
<Fallenou>
ok
<Fallenou>
thanks
<lekernel>
the NOR image would indeed be YAFFS2, but right now there is no way to create such an image externally
<Fallenou>
it looks like there is some doc about how to integrate yaffs2 on linux 2.6 kernel
<lekernel>
yeah, but you'd run into the OOB data problem when trying to use a normal image
<Fallenou>
so the fs  is created at boot time ?
<lekernel>
the FS port I did for RTEMS moves the OOB data in-band
<Fallenou>
oh
<lekernel>
you'd need to use the same OOB data layout in Linux to get that to work
<lekernel>
needs a bit of coding
<Fallenou>
hum ok
<Fallenou>
so it's a special version you have in rtems
<Fallenou>
you put the signaling with the data
<lekernel>
the yaffs core passes two pointers to the read/write functions, one is for normal data and the other for oob
<lekernel>
so you can mix the two without hacking all over the yaffs2 code
<Fallenou>
hum I did a mkfs.vfat on my sdcard file
<Fallenou>
mounted it via -o loop
<Fallenou>
put a file on it
<Fallenou>
booted qemu with -sdcard myfile
<Fallenou>
(after unmounting it
<Fallenou>
where should I see it in rtems ?
<Fallenou>
the /flash only contains lost+found
<lekernel>
yeah
<lekernel>
this folder is always present on yaffs, even with a blank flash
<lekernel>
the memory card is mounted in /memcard
<lekernel>
it should have a partition table too
<lekernel>
you'll probably need to mess a bit with fdisk and dd to generate an image that has both a partition table and a filesystem
<Fallenou>
hum yes ok i just did a partition but no partition table
<Fallenou>
humm sending files via ftp from ubuntu to rtems sucks :(
<lekernel>
or you could mount it without a partition table, using the rtems shell
<Fallenou>
but the other way it works
<lekernel>
but it won't be automounted
<Fallenou>
oh !
<Fallenou>
I can do that
<Fallenou>
but which /dev/flash* to use ?
<lekernel>
it's /dev/memcard
<lekernel>
/dev/flash* are the different NOR partitions
<Fallenou>
great it mounted
<Fallenou>
it has my file on it
<Fallenou>
downloading the flickernoise binary from rtems
<Fallenou>
via ftp
<Fallenou>
I will test a md5sum on it after :)
<Fallenou>
transfered 9 MB in 47 sec, not really fast for a loopback connexion :'
<Fallenou>
aouchhh
<Fallenou>
not the same md5sum
<Fallenou>
tries again with a smaller file
<lekernel>
if it's for testing ftp, you can try the ramdisk
<Fallenou>
ok
<Fallenou>
how can I umount /mnt ?
<Fallenou>
oh ok it's unmount
<lekernel>
it removes the memory card related sources of potential data corruption
<Fallenou>
ok i've done the same on a small 10 Ko file the md5sum is the same
<lekernel>
reminds me of my multi-dozen-terabyte memory controller testing fun :)