00:00
luka177 has joined #linux-amlogic
00:12
luka177 has quit [Ping timeout: 260 seconds]
00:49
GNUtoo has quit [Ping timeout: 252 seconds]
00:51
GNUtoo has joined #linux-amlogic
02:19
hexdump0815 has quit [Ping timeout: 245 seconds]
02:21
hexdump0815 has joined #linux-amlogic
06:19
djrscally has joined #linux-amlogic
06:43
naoki has joined #linux-amlogic
06:51
konsgn has quit [Ping timeout: 252 seconds]
06:51
konsgn9 has joined #linux-amlogic
07:28
knuxify2 has joined #linux-amlogic
07:40
paulk has quit [Ping timeout: 245 seconds]
07:54
paulk has joined #linux-amlogic
07:54
paulk has joined #linux-amlogic
08:21
ldevulder has joined #linux-amlogic
08:28
ldevulder has quit [Quit: Leaving]
09:04
jacobk has quit [Ping timeout: 248 seconds]
09:05
jacobk has joined #linux-amlogic
11:13
konsgn has joined #linux-amlogic
11:15
konsgn9 has quit [Ping timeout: 276 seconds]
11:24
psydroid has joined #linux-amlogic
11:37
naoki has quit [Quit: naoki]
13:29
psydroid2 has joined #linux-amlogic
13:33
chewitt has joined #linux-amlogic
14:30
buzzmarshall has joined #linux-amlogic
15:03
konsgn8 has joined #linux-amlogic
15:05
<
lvrp16 >
the new spicc dma patch does not send valid data on certain sizes like the old PIO code.
15:05
konsgn has quit [Ping timeout: 252 seconds]
15:05
konsgn8 is now known as konsgn
15:06
<
lvrp16 >
some of the SPI_BURST_LEN logic does not look like it makes sense, I'll try to fix/rewrite it and try to test if I can shoehorn in DMA for small words in.
17:20
chewitt has quit [Quit: Zzz..]
17:39
vagrantc has joined #linux-amlogic
19:17
<
f_ >
first time using b4 :P
20:24
<
lvrp16 >
well, dma does not work on GXL. I've built a test suite for SPI and some certain speeds and bpw combinations don't work. spi is a mess.
20:24
<
lvrp16 >
the current driver needs to be overhauled. will see if I finish it this weekend.
20:26
<
lvrp16 >
and certain size transfers don't work either with the new dma code even though it should. 4136 bytes for example
20:27
<
lvrp16 >
running an exhaustive test against every 8 byte transfer from 16B up to 1MB on GXL, G12B and SM1 right now, for 8 16 32 64 bpw.
21:04
vagrantc has quit [Ping timeout: 265 seconds]
21:47
f_[x] is now known as funderscore
22:01
funderscore is now known as f_1
22:14
<
f_ >
Ok so I'm wondering if I should upstream my tv stick u-boot patches after the DT patches..
22:15
<
f_ >
that device is special in that regular amlogic bl2 will not work
22:15
<
f_ >
because secureboot is enabled. So one needs to either patch bl2 to remove secureboot checks, or use the wip u-boot-spl stuff
22:16
<
f_ >
speaking of U-Boot SPL I've been cleaning up the code lately
22:33
<
lvrp16 >
wait, bl2 can be patched to remove secureboot?
22:33
<
lvrp16 >
so for GXL, it's wide open?
22:34
<
f_ >
lvrp16: you can do binary patching to remove BL2's secureboot checks once you gain unsigned code execution yeah.
22:34
<
f_ >
either that or run u-boot-spl
22:34
<
lvrp16 >
so there's no chain of trust from bl1 to bl2?
22:34
<
lvrp16 >
eg, the bootrom does not verify bl2 signature?
22:34
<
f_ >
BL1 does checks to BL2
22:35
<
lvrp16 >
then how do you patch bl2 and still have the sig be valid?
22:35
<
f_ >
when secureboot is enabled it checks BL2's signature. It can also decrypt BL2 if it's AES-encrypted and the efuses have been burned to allow it
22:35
<
f_ >
lvrp16: this is where the amlogic-usbdl vuln comes in
22:36
<
f_ >
^ because of this, you can run unsigned code from the bootROM's USB mode
22:36
<
f_ >
It's inconvenient to do, but allows you to do all sorts of fun stuff
22:37
<
f_ >
like, run your own BL2 payload!
22:37
<
f_ >
at this point the payload can be anything. So it can be u-boot-spl, just like it can be a BL2 patched to remove secureboot checks
22:38
<
f_ >
The only way to mitigate it, is to lock USB mode behind a password.
22:38
<
f_ >
and this affects almost all pre-2021 Amlogic SoCs, to my knowledge
22:39
<
lvrp16 >
ahh did they ever patch bl1?
22:39
<
f_ >
Not to my knowledge, no
22:40
<
f_ >
My TV Stick's BL1 is identical to lepotato's BL1 which is identical to lafrite's BL1
22:40
<
f_ >
so you can be certain all three have the vulnerability
22:40
<
lvrp16 >
Well, that site said S905D3
22:40
<
f_ >
yes, it was originally found on S905D3
22:41
<
lvrp16 >
yeah so I'm wondering if they ever patches S905D3 and A311D bl1 since these chips are still active.
22:41
<
f_ >
I could probably dump alta's ROM to compare
22:43
<
f_ >
or, well, dump solitude's ROM too since that one has S905D3
22:50
<
f_ >
If they patched it, could be worth documenting.
22:52
djrscally has quit [Ping timeout: 265 seconds]