**** BEGIN LOGGING AT Sun Mar 28 02:59:56 2021 Mar 28 03:13:36 thank you! I'll look into that Mar 28 04:21:38 frosteyes: https://github.com/MentorEmbedded/meta-mentor/blob/master/meta-mel-support/classes/image-sanity-incompatible.bbclass may be of use to you, if you set INCOMPATIBLE_LICENSE globally to a value that allows gplv3 to be built, but set it differently in the image, and use this class, it'll error out if someone tried to install a gplv3 package into the image, but would allow it to build as usual Mar 28 04:22:09 frosteyes: we had a case where we whitelisted something to let it build, but never wanted it to be installed in a production image, only a development one, and this made sure that was the case for us Mar 28 08:13:19 Thanks kergoth it looks very much what I need Mar 28 08:36:42 alex88: to create ipk's, there is the command bitbake -c do_package_write_ipk , but that's all hacky. I recommend the life coding sessions with LetoThe2nd to get an idea of the yocto workflow :) : https://www.youtube.com/watch?v=ThTl4FItfMI&list=PLD4M5FoHz-TxMfBFrDKfIS_GLY25Qsfyj Mar 28 16:43:44 Hi, I'm looking to install ext sdk generated by yocto under WSL2 but getting a :/mnt/c/Program: bad variable name from post-relocate-setup.sh, not so critical for regular sdk since it doesnt prevent actual sdk from functioning, but for sdk_ext it blocks it completely Mar 28 16:43:44 Is anybody aware of a workaround? I understand that it's a clash of windows/linux folder naming and WSL2 is not supported, but maybe somebody already encountered this issue? Mar 28 19:25:33 Hi folks. Has anyone worked with squashfs, and had issue with performance? Mar 28 19:27:08 I have some qspi nor flash (in theory it would be able to do 150 MB/s) and if I use dd directly on the /dev/mtd device it perform around 114 MB/s. So pretty good Mar 28 19:28:03 I have then created a mtd partition of 110 MB and created a squashfs using xz with block size of 1M. written it to flash and mounted it.. Mar 28 19:28:27 (it only contains a single file currently). When I try to read this file with dd, it only gives around 6.2 MB/s. Mar 28 19:28:53 Knows that the decompression takes some overhead, but 6.2 MB/s is much slower than I would expect. Mar 28 19:29:11 So any input is welcome :D Mar 28 20:18:27 Does a bigger dd blocksize help? Maybe trying another SquashFS compressor? Mar 28 20:30:26 I have added kprobe tracing with echo 'p squashfs_read_data super_block=%x0 index=%x1 length=%x2' > kprobe_events Mar 28 20:31:48 It gives the following result. Mar 28 20:31:54 dd-2141 [001] d... 9134.666453: p_squashfs_read_data_0: (squashfs_read_data+0x0/0x5c8 [squashfs]) super_block=0xffffffc9737eb000 index=0x60 length=0x48c74 Mar 28 20:31:56 dd-2141 [002] d... 9134.821142: p_squashfs_read_data_0: (squashfs_read_data+0x0/0x5c8 [squashfs]) super_block=0xffffffc9737eb000 index=0x48cd4 length=0x4b830 Mar 28 20:31:58 dd-2141 [001] d... 9134.988574: p_squashfs_read_data_0: (squashfs_read_data+0x0/0x5c8 [squashfs]) super_block=0xffffffc9737eb000 index=0x94504 length=0x4ae4c Mar 28 20:33:49 So length fluctuate and does not seems to fit with block - on this level.. Mar 28 20:34:24 Will just see if I can find the place where the data from the mtd is requested.. So I can see what size is requested.. Mar 28 21:13:23 Using kprobe I noticed that all mtd_read was 0x200 bytes (512) Mar 28 21:13:51 This is the mtdblock device, where I can see that .blksize is set to 512 in mtdblock.c Mar 28 21:30:02 Makes sense so far. 1M block size means mksquashfs chops the input up into 1M blocks that it then compresses to variable sized output blocks. And of course you need a block device wrapper underneath, because SquashFS cannot talk to UBI or MTD. Mar 28 22:47:14 goliath: I was not prepared on the slow mtdblock, and I can't change blocksize.. Mar 28 22:47:53 ~ # dd if=/dev/mtd3 of=/dev/null bs=128k count=10 Mar 28 22:47:55 1310720 bytes (1.3MB) copied, 0.011095 seconds, 112.7MB/s Mar 28 22:47:57 ~ # dd if=/dev/mtdblock3 of=/dev/null bs=128k count=10 Mar 28 22:47:59 1310720 bytes (1.3MB) copied, 0.362195 seconds, 3.5MB/s Mar 28 22:48:33 So will need another way in using the mtdblock. Mar 28 22:49:50 Will use the kernel mailing list, but just a quick question. Do you know if UBI is able to work with locked flash, or does UBI require to be able to write to the flash. Mar 28 22:50:48 The reason I ask, is for this part, it is production data so is readonly, and I will prefer to be able to lock the flash in production after the data is written. Notice it is nor flash. Mar 29 02:56:12 * paulg wonders who'd complain if we declared xxd-native a part of ASSUME_PROVIDED... **** ENDING LOGGING AT Mon Mar 29 02:59:56 2021