| 09:37 | vpeter> | jnettlet: one quick question if I may. I have i2c expander with gpio outputs and connected led. Is it possible somehow to declare this led directly in device tree to use it like /sys/class/leds/red ? |
| 09:52 | jnettlet> | vpeter: yep of course |
| 09:55 | jnettlet> | well. let me preface that. As long as the expander driver registers itself to the OF GPIO infrastructure |
| 10:31 | vpeter> | Ok, then I need to google little more :) |
| 10:41 | vpeter> | Ah, seems PCA953x doesn't register itself :( |
| 12:42 | Exaga> | jnettlet: :> |
| 12:42 | Exaga> | hi vpeter |
| 12:43 | Exaga> | jnettlet: i have just attended a job interview and if i get the position... who sells the honeycomb? :D |
| 12:43 | Exaga> | hahaha |
| 12:44 | jnettlet> | Exaga: Good Luck! Did the interview go well? |
| 12:46 | Exaga> | jnettlet: it was for a centos admin |
| 12:46 | Exaga> | so quite easy |
| 12:46 | Exaga> | the questions they asked me i answered them all except "explain what an inode is?" |
| 12:47 | Exaga> | i replied "an inode is a name for something which has always existed that I don't know the description of" |
| 12:47 | Exaga> | then they told me it contained group permissions... so i instructed them how to view/change/add/remove file and directory permissions without using the term "inode" :> |
| 12:48 | jnettlet> | I don't know many sysadmins that need to know what an inode is at this point. |
| 12:48 | EmilKarlson> | inodes are mostly relevant for ext4 inode tables, where you can actually run out of inodes |
| 12:48 | jnettlet> | well it contains more than just the group permissions...it is a basic file system structure |
| 12:48 | Exaga> | jnettlet: if u were my boss and said to me "user12345 needs adding to the 'trusted' group" i'd usermod +aG trusted user12345 |
| 12:49 | EmilKarlson> | so you'll need to know how to format ext4 for these special use cases, where you have loads of small files |
| 12:49 | Exaga> | mkfs.ext4 /dev/filesystem |
| 12:49 | jnettlet> | or use XFS which is what most EPEL distributions use by default now :) |
| 12:50 | EmilKarlson> | yeah XFS is really strong for traditional FS use cases |
| 12:50 | Exaga> | cool EmilKarlson |
| 12:50 | EmilKarlson> | I used XFS on talos2, since it had really good support for both 4K and 64K page sizes with same on disk specs |
| 12:51 | jnettlet> | yep, I use it on most my server type machines. |
| 12:51 | Exaga> | people tend to use phrases and term which mean nothing only to those concerned |
| 12:51 | jnettlet> | started playing around with f2fs on embedded, they are making nice strides there |
| 12:52 | Exaga> | jnettlet: i bet you can reel off acronyms like there's no tomkorrow |
| 12:52 | jnettlet> | plus for a sysadmin most these things are easily googlable on the web |
| 12:52 | EmilKarlson> | but I would say inode is basically a file excluding the parts of the files that link it to directories, like file names |
| 12:52 | Exaga> | EmilKarlson: that's what google told me |
| 12:52 | EmilKarlson> | but it's a bit complicated, as filesystems have varying concepts of on-disk inodes |
| 12:53 | EmilKarlson> | like microsoft filesystems do not have them as such |
| 12:53 | jnettlet> | yeah way too much jargon to remember |
| 12:53 | Exaga> | EmilKarlson: inodes are linux |
| 12:53 | jnettlet> | and Mac has their own variant |
| 12:53 | EmilKarlson> | yes, they do have ls -i support with inode numbers |
| 12:54 | Exaga> | and yes an inode is a data structure on linux or unix that stores the info about a file except for it's name and any data it contains |
| 12:54 | EmilKarlson> | apfs has similar inode number as b-tree key as btrfs has |
| 12:54 | Exaga> | so that leaves file permissions and little else? |
| 12:54 | EmilKarlson> | well again, depends on the filesystem |
| 12:55 | EmilKarlson> | but linux vfs has struct inode, which is same for all filesystems for the in-memory part |
| 12:55 | Exaga> | ahhh |
| 12:55 | Exaga> | now i get it |
| 12:55 | EmilKarlson> | typically it's used as a member of filesystem specific struct |
| 12:56 | Exaga> | EmilKarlson: you're better than google. thank you |
| 12:56 | Exaga> | yes |
| 12:56 | Exaga> | now it makes sense |
| 12:56 | EmilKarlson> | there is a bit of complexity as there are different namespaces for different levels |
| 12:57 | Exaga> | different filesystems use inodes in different ways but they are still inodes |
| 12:57 | EmilKarlson> | like there is userspace api (see ls -i), vfs and in memory structures and on-disk structures |
| 12:57 | Exaga> | like spoken languages, they all have bread and water but call it by colloquial names |
| 12:57 | EmilKarlson> | well for userspace api they always have to have inode number |
| 12:57 | EmilKarlson> | on-disk many do not |
| 12:57 | Exaga> | EmilKarlson: that's way over my head lol |
| 12:58 | Exaga> | jnettlet knows what you're talking about. i dont lol |
| 12:58 | Exaga> | anyway... forget about the job and inodes. HONEYCOMB! |
| 12:58 | Exaga | 12:58 * Exaga needs a job so he can buy one |
| 12:58 | EmilKarlson> | it all used to be simpler, when there was just the unix filesystem where everything was mapped one to one on all layers |
| 12:58 | Exaga> | it's like the 3rd of 4th thing on my list |
| 12:59 | EmilKarlson> | most filesystems are no longer fully posix in all aspects |
| 12:59 | Exaga> | hmmm |
| 13:00 | Exaga> | before the 2nd month's payday... that's the plan if i get the job |
| 13:00 | Exaga> | then i'll start building the contrete bunker in the back garden |
| 13:00 | Exaga> | to store it in |
| 13:01 | Exaga> | concrete* |
| 13:03 | Exaga> | jnettlet: is the honeycomb RAM-specific or will it take most DDR4 modules? |
| 13:04 | jnettlet> | Exaga: it should take most DDR4 SODIMMS. With the latest BSP it can also handle running with a single DIMM, so you can expand later. |
| 13:04 | jnettlet> | however we are still doing compatibility testing. I just picked up a pair of Crucial ECC modules to test, although they are expensive and a bit hard to come by. |
| 13:05 | jnettlet> | the fastest I could find were 2666 MT/s |
| 13:05 | EmilKarlson> | if only mcbin was not sufficient, I could buy that as well |
| 13:06 | EmilKarlson> | I need more java workloads |
| 13:06 | EmilKarlson> | well, I could start hosting a matrix homeserver at home |
| 13:07 | jnettlet> | Multi-core java workloads really benefit from the large L3 cache |
| 13:07 | Exaga> | jnettlet: i use HyperX and Samsung usually |
| 13:07 | Exaga> | but whatever works :> |
| 13:08 | jnettlet> | HyperX 3200 is on my list |
| 13:09 | EmilKarlson> | tbh. I don't think I have fast enough network for matrix homeserver |
| 13:11 | Exaga> | jnettlet: does SolidRun/atai/rabeeh/kossay need a teaboy yet??? |
| 13:11 | Exaga> | tell them i make the best cup of tea ever! |
| 13:11 | Exaga> | and i can part-time as rabeeh's personal hardware tester |
| 13:12 | Exaga> | i'll do that for free, but they'll have to pay me a salary to make tea |
| 13:14 | jnettlet> | no tea in the office. You will need to learn to make Arabic coffee :) |
| 13:14 | Exaga> | ok then give me some time to grep |
| 13:14 | Exaga> | man arabic-coffee |
| 13:16 | Exaga> | hmmm: /configure |
| 13:23 | Exaga> | make -j3 PATH=/root/directors_office ARCH=לְהַרְווֹת LOCALVERSION="קפה_ערביקה" CROSS_COMPILE=Qahwah-arabiyya-gcc bzArabica-coffee |
| 13:31 | Exaga> | cp -av arch/arm64/boot/bzArabica-coffee /mnt/Solidrun/office_kitchen/coffee_bean_storage_jar/ |
| 13:32 | Exaga> | jnettlet: where does SolidRun keep their finjan and della? :> |
| 13:35 | jnettlet> | In the new Cafe / Kitchen that was just built out last time I was there |
| 13:36 | Exaga> | jnettlet: whereis cafe && whereis kitchen |
| 13:36 | Exaga> | find / -name "cafe" |
| 13:36 | Exaga> | which "inode" is it in? hahaha |
| 13:50 | Exaga> | jnettlet: on a serious note, from what i've read on arabic coffee it's fascinating |
| 13:50 | jnettlet> | I don't much about the history. It is very strong, I know that for sure |
| 13:51 | Exaga> | it's a middle-eastern protocol |
| 13:51 | Exaga> | very impressive |
| 13:51 | Exaga> | i had no idea about this. thanks for sharing |
| 13:52 | Exaga> | what an art it is and such skill involved |
| 13:52 | Exaga> | care and attention to detail too |
| 13:59 | jnettlet> | yep it is quite a process |
| 14:11 | Exaga> | yes and thanks jnettlet i will certainly be educating myself with this culture |
| 14:11 | Exaga> | proves my point that Solidrun are more than great innovators. they are great teachers too! |
| 17:45 | vpeter> | jnettlet: Could I borrow you again? What do I need to do to have red_led actually as led? |
| 17:45 | vpeter> | # cat /sys/kernel/debug/gpio |
| 17:45 | vpeter> | gpiochip2: GPIOs 402-409, parent: i2c/4-0020, tca6408, can sleep: |
| 17:45 | vpeter> | gpio-402 (p0 ) |
| 17:45 | vpeter> | gpio-403 (p1 ) |
| 17:45 | vpeter> | gpio-404 (p2 ) |
| 17:45 | vpeter> | gpio-405 (p3 ) |
| 17:45 | vpeter> | gpio-406 (p4 ) |
| 17:45 | vpeter> | gpio-407 (red_led ) |
| 17:45 | vpeter> | gpio-408 (p6 ) |
| 17:45 | vpeter> | gpio-409 (p7 ) |
| 17:46 | jnettlet> | vpeter: just turning on the LED from the command line, or integrating it into device-tree? |
| 17:47 | vpeter> | Integrating into device tree. I can control it manually from user space by exporting gpio, set direction, writing vaue to it. |
| 17:48 | vpeter> | Would like to make "nicer" approach :) |
| 17:57 | jnettlet> | vpeter: https://www.kernel.org/doc/Documentation/devicetree/bindings/leds/leds-gpio.txt |
| 17:58 | jnettlet> | it is pretty straight forward |
| 18:11 | vpeter> | Working! Thank you. Didn't understand how to specify gpio at first. |
| 18:18 | jnettlet> | great |
| 18:20 | vpeter> | You are THE man! |