<ssvb>
jemk: I'll try to run some more tests and will probably revert it
<ssvb>
jemk: right now I'm about 90% sure that it is the root cause
<ssvb>
jemk: I have also tried some experiments with SDR_ZQCR1
<ssvb>
jemk: it is indeed responsible for the periodic re-calibration at runtime, because I see that the value in SDR_ZQSR keeps changing
<ssvb>
jemk: however the reliability gets much worse
<ssvb>
jemk: now I can only suspect that it is doing some "short" periodic calibration, which has poor accuracy compared to the proper "long" calibration at the initialization time
<ssvb>
jemk: maybe enabling burst refresh can improve something (give the zq calibration a bit more time to do its job?)
vinceh_ has quit [Ping timeout: 256 seconds]
issue_ has joined #linux-sunxi
<jemk>
ssvb: I have 0x10d1800 after coldstart and still have it now after hours of lima-memtester with reboots
<ssvb>
jemk: you can try to set zq to 0x10d3900 and check if something changes
<ssvb>
jemk: this will bypass the calibration and just use this value directly
<jemk>
ssvb: and then scan tpr3 again?
<ssvb>
jemk: yes, if you have time for this experiment
<jemk>
ssvb: should be possible, will need less time than hardening tests ;)
<ssvb>
jemk: one more interesting thing in your html report is that the read errors seem to be concentrated in one lane, and write errors are concentrated in another one
<jemk>
ssvb: yep, i noticed that too, and read errs are top right, write bottom left
<ssvb>
trying per-lane delay adjustments may probably provide a good improvement now
<ssvb>
earlier I was testing per-lane delay adjustments, but that was before I realized that the impedance mismatch is much more important
<ssvb>
this can be revisited again
vinceh_ has joined #linux-sunxi
Gerwin_J has joined #linux-sunxi
hypno__ is now known as hypno
vinceh_ has quit [Ping timeout: 255 seconds]
HeHoPMaJIeH has quit [Quit: Leaving]
tomboy65 has quit [Remote host closed the connection]
avsm has joined #linux-sunxi
tomboy65 has joined #linux-sunxi
dack has joined #linux-sunxi
jemk has quit [Quit: Lost terminal]
xavia has joined #linux-sunxi
vinceh_ has joined #linux-sunxi
FreezingCold has quit [Ping timeout: 250 seconds]
jemk has joined #linux-sunxi
issue_ has quit [Remote host closed the connection]
prahal has joined #linux-sunxi
mmarker has quit [Read error: Connection reset by peer]
mmarker1 has joined #linux-sunxi
FreezingCold has joined #linux-sunxi
prahal has quit [Read error: Connection reset by peer]
prahal has joined #linux-sunxi
blsd has quit [Remote host closed the connection]
xinj has quit [Ping timeout: 264 seconds]
focus has quit [Ping timeout: 250 seconds]
TheSeven has quit [Ping timeout: 272 seconds]
libcg has quit [Remote host closed the connection]
<libv>
dack: yes, but you do not want to know how many hours i burned on that
<dack>
libv: ^_^
<dack>
libv: okay, but is the only way to figure out why it's not working is with a UART connection?
<libv>
i am currently writing a small patch to make the scan fail more gracefully
<dack>
libv: gracefully? You mean with logging? Or are you talking about the nandd process going to 100% CPU?
<libv>
so that it actually complains about not finding the nand and prints out a chip id
<dack>
libv: with the module in-built module I get messages saying "Dev Sunxi softw411 nand magic does not match for MBR 1: <FF><FF><FF><FF><FF><FF><FF><FF>"
paulk-collins has quit [Remote host closed the connection]
<libv>
hexdump -C it and pastebin that
<dack>
libv: k.. give me a few minutes. I switched back to the module to save CPU
<dack>
libv: if I "modprobe nand" it seems to load, but nothing happens
<libv>
nand needs an endless amount of work
<libv>
someone should actually go play with the cubietruck and the new partition scheme of the more recent cubie images
<libv>
but the amazing thing is, i either go do that myself, or people will just keep on complaining/crying
<libv>
and if i do that, people will complain that i don't work on kms
<libv>
dack: perhaps you can solve this for us.
<dack>
libv: my skills are limited...
<dack>
libv: is the "SUNXI Nandflash Test Driver" useful for this? I'm not sure what it does
<libv>
i have no idea what that is
<libv>
hdmi on a20 needs a tiny bit more init before hpd works...
physis has joined #linux-sunxi
<dack>
libv: I'm guessing the 1Mb dd should only take a second, right? It seems to be frozen...
kivutar has joined #linux-sunxi
<ijc>
ssvb: I see your DRAM series, good work. I've only glanced through it so far and although I'm not going to pretend I understand it all it looks pretty good.
<ijc>
I think a change of that magnitude needs to sit on the list for a bit before being applied so people have a chance to comment etc, so it's not part of the next PR I sent just now.
<dack>
libv: sending USR1 to dd won't even give me a status on the copy...
<jemk>
ssvb: tpr3-scan is still running, but I already have a much bigger green area with the zq fixed at 0x10d3900 (autodetect was 0x10d1800)
<libv>
dack: then it's time to start instrumenting the nand driver
<Turl>
dack: dd likes to print them when it finishes a block or so, if you used a big bs= it may wake a while
<libv>
Turl: for 1MB?\
<Turl>
1MB should be done in a couple seconds tops
<libv>
that has SCAN_DBG("[SCAN_DBG] Nand flash chip id is:0x%x 0x%x 0x%x 0x%x 0x%x 0x%x\n",
<libv>
turn that into a pr_info and recompile, and see the id printed out
<dack>
libv: sorry.. I'm not sure how to do that. pr_info?
<libv>
well, i am not going to do that for you.
<dack>
libv: no, no.. I'm not sure how to get that debug statement to print out or what you mean by "pr_info"
<libv>
google it
<Turl>
dack: replace SCAN_DBG( with pr_info( and recompile
<Turl>
that's all there is to it
libcg has quit [Ping timeout: 255 seconds]
<dack>
Turl: k.. then it should come out in syslog?
<Turl>
dmesg, yeah
<dack>
Turl: thanks..
<libv>
*sigh* spoonfeeding.
<Turl>
yw
<Turl>
libv: there's barely anything to feed
<libv>
and still you let him have that
<Turl>
it's not like I'm going over the compilation
<dack>
libv: I'm not that familiar with kernel coding...
<libv>
dack: and?
<dack>
libv: I've never heard of pr_info()
<libv>
then why didn't you just look it up?
<dack>
libv: I did.. but Turl saved me 30 minutes of reading just to figure out it works like printf
<libv>
3 minutes, tops.
<dack>
libv: any way.. how would you get SCAN_DBG() printing out?
<libv>
by finding the define that turns it into something useful, but that then enables all scan debug printing or even worse
<libv>
forest, trees and such
<libv>
again, this is very basic stuff
<Turl>
libv: talking about kms..
<Turl>
libv: do you recall how is hdmi audio handled?
<libv>
Turl: sent out in the blanking
<Turl>
libv: I mean in sunxi, is it via iis?
<libv>
how it gets to the hdmi block?
<libv>
no idea, never tried it
<libv>
i am also not intending to implement it, there are tons of people peeing their pants to do it
<dack>
libv: see that's another thing I'm not familiar with... developing by peeing one's pants
<dack>
I'm learning so much. ^_^
<libv>
dack: hdmi cec and hdmi audio are things some people seem overly interested in
<libv>
dack: people are not so interested in the more boring bits
<libv>
peeing in pants is being overexcited, like a puppy
<dack>
libv: it was a joke.. I understood what you were saying
<dack>
libv: nand flash chip id is: 0xad 0xde 0x94 0xeb 0x74 0x44
<libv>
hrm, that should match
<libv>
according to nand_id.c
<libv>
so that cannot be the issue
<dack>
libv: really? the nand_id.c shows 0xff instead of 0x44, no?
<dack>
libv: or is that a mask?
<libv>
that most likely tells the code to ignore this value in id-ing the chip
<libv>
so that's not it, size matches as well
forcev has quit [Ping timeout: 264 seconds]
<dack>
libv: I seem to be able to access it with a terminal in Android... but files like script.bin seem to be missing.. maybe some sort of corruption with the data? (though everything seems to run fine in Android)
FunkyPenguin has joined #linux-sunxi
kz1 has quit [*.net *.split]
montjoie[home] has quit [*.net *.split]
montjoie[home] has joined #linux-sunxi
focus has quit [*.net *.split]
ecelis has quit [*.net *.split]
egbert has quit [*.net *.split]
slapin_ has quit [*.net *.split]
Nyuutwo has quit [*.net *.split]
Zboonet has quit [*.net *.split]
ccaione has quit [*.net *.split]
mtr has quit [*.net *.split]
ecelis has joined #linux-sunxi
libcg has joined #linux-sunxi
egbert has joined #linux-sunxi
focus has joined #linux-sunxi
Zboonet has joined #linux-sunxi
kz1 has joined #linux-sunxi
Nyuutwo has joined #linux-sunxi
<libv>
ok, those pastebinned things are just warnings from the partition checking code
slapin_ has joined #linux-sunxi
mtr has joined #linux-sunxi
<libv>
heh. why?
<libv>
why is everything and the kitchen sink included in sun7i_defconfig
<libv>
except nand.
FunkyPenguin has quit [Ping timeout: 240 seconds]
<libv>
dack: first off, if built as a module, it seems to be called sunxi_nand
<dack>
libv: isn't it included, but as a module?
akaizen_ has quit [Remote host closed the connection]
<libv>
nope, not in defconfig
<libv>
CONFIG_SUNXI_NAND
FunkyPenguin has joined #linux-sunxi
akaizen has joined #linux-sunxi
kivutar has quit [Read error: Connection reset by peer]
<dack>
libv: yeah.. I see it as a module after I do "make ARCH=arm sun7i_defconfig"
dreamer has quit [Remote host closed the connection]
physis has quit []
<dack>
libv: then I switch SUNXI_NAND to built-in
<libv>
ah, since march 2nd
<libv>
dack: when you modprobe "nand", nothing will happen
<libv>
you have to modprobe sunxi_nand
<libv>
after which time you should see the exact same happen as with built in, only at a different time
<dack>
libv: I think I tried that first... it can't find it. If I do "nand" than it shows something in lsmod as being loaded
<libv>
because nand is a completely different thing
<dack>
libv: ah.. :) of course.
<libv>
pfff
kivutar has joined #linux-sunxi
<dack>
libv: which .ko is supposed to be sunxi-nand?
akaizen has quit [Ping timeout: 240 seconds]
<libv>
-?
<libv>
really?
<dack>
libv: I'll recompile and see if I can find it... I don't see anything there.
<libv>
anyway, my tree is too old for this, i'd have to recompile too much for this, i don't have enough patience for this anyway, i'm out
<libv>
oh, and it is seriously in the way of me implementing ddc for hdmi for kms
<dack>
there's nothing remotely "sunxi-nand" in kernel/drivers/block/
libcg has quit [Remote host closed the connection]
libcg has joined #linux-sunxi
<libv>
oh, it is called nand, as well
<libv>
that means that there are two nand modules?
Black_Horseman has joined #linux-sunxi
Black_Horseman has joined #linux-sunxi
<Black_Horseman>
Hell -o
Deskwizard has quit [Quit: No Ping reply in 180 seconds.]
Gerwin_J has quit [Quit: Gerwin_J]
Deskwizard has joined #linux-sunxi
<dack>
libv: at the risk of being chastised.. where is the final module name defined?
<libv>
Makefile
<dack>
libv: which one? :)
<libv>
drivers/block/sunxi_nand/nfd/Makefile
<libv>
which is spoonfeeding again
<dack>
libv: I thought the name you passed to modprobe usually matched the .ko file name... why is there no nand.ko?
<dack>
kernel/drivers/mtd/nand/nand.ko
<dack>
found it..
<dack>
or is that the "other one"
<dack>
I have a kernel/drivers/block/nbd.ko ...
ccaione has joined #linux-sunxi
<dack>
nope.. that's a network block device..
akaizen has joined #linux-sunxi
<dack>
libv: When I do "make ARCH=arm CROSS_COMPILE=arm-linux-gnueabihf- -j4 INSTALL_MOD_PATH=output modules_install" I don't think the proper sunxi_nand modules are copied over... there's something wrong in a Makefile somewhere
<oliv3r>
wens: mipi alliance says CSI stands for Camera Serial Interface, and CPI for Camera Parallel Interface :)
gzamboni has quit [Read error: Connection reset by peer]
libcg has quit [Quit: rtfm]
akaizen has quit [Remote host closed the connection]
FunkyPenguin has quit [Ping timeout: 245 seconds]
FunkyPenguin has joined #linux-sunxi
<dack>
libv: okay, after disabling the "other nand" module then I now have a kernel/drivers/block/sunxi_nand/nfd/nand.ko
<dack>
libv: with the current defconf there's no way anyone could be using sunxi_nand
<libv>
dack: time to make a patch and send it in to the ml
deasy has quit [Quit: Nom d'un quark, c'est Edmonton !]
<dack>
libv: is it just changing "obj-$(CONFIG_SUNXI_NAND) += nand.o" to be something like sunxi_nand.o instead?
FunkyPenguin has quit [Ping timeout: 240 seconds]
<libv>
dack: or fixing defconfig, whichever you wish to do
<dack>
libv: hmm.. I have no idea what that other nand.ko even does
FunkyPenguin has joined #linux-sunxi
phase1_ has joined #linux-sunxi
gzamboni has joined #linux-sunxi
<phase1_>
hello, i have a cubietruck a20, and was wondering if it was possible to change the speed of the i2c bus. i want to enable fast mode. if so, would it be through code or a system variable?