freenode/#mezzano - IRC Chatlog
Search
20:20:55
fittestbits__
froggey: I have a collection of reference documents: USB specs and FAT spec. I'm using them for writing code for mezzano.
20:21:19
fittestbits__
I'd like to make them available on github so that other people working on mezzano can reference them.
20:22:35
fittestbits__
But, I was wondering if you would like them in the mezzano tree perhaps under doc/ref?
20:28:36
fittestbits__
Yeah, me too. So, not in the mezzano tree. Thanks! I'll have to think about if I really want to create a repository for them ...
20:41:00
random-nick
well, does their license allow redistribution? if not it's probably not a good idea
20:58:37
fittestbits__
Looks like they have a standard copyright notice. So, no license allowing redistribution. It seems like a weird case to me though, since they are freely available for download.
20:59:40
fittestbits__
I guess I should try to figure out which URL I used to download them - I didn't keep track. A list of URLs would be safe, right?
21:10:37
fittestbits__
I'll see if I can find URLs for the copyright holders. At then I can give those out if anyone is interested.
21:19:12
fittestbits__
Oh, maybe that's it. ELS 21 means 2021! I didn't get that from the notice. The web site says 2020 in Zurich. End of April.
1:02:33
fittestbits__
yes - was just working on recognizing if a partition or disk is a fat file system.
1:11:52
fittestbits__
Problem is block-device-read-sector reads some number of sectors, so data in hash table may not be the right size - too small or too large.
1:12:35
fittestbits__
So, do you cache individual sectors and merge copy them into the right size buffer based on the request.
1:13:50
fittestbits__
OK, but you still have to match the data in the cache against the request, right?
1:17:05
fittestbits__
OK, so if block size is 4KB, then address of the operation must also be 4KB aligned so that the cache blocks line up with the request.
1:21:23
ebrasca
Now how it is done first request to some start-sector with some size is what next request to place start-sector is going to get.
1:24:01
fittestbits__
I think it would be better if the callers didn't have to know what is going on in the cache.
1:25:11
fittestbits__
So, I would cache sector size buffers, then either copy the sectors from the cache to the buffer or read data from the disk into the buffer as required.
1:25:55
fittestbits__
Yes, which is a disadvantage. Of course memory to memory copy is much faster than disk I/O.
1:27:06
fittestbits__
It would be better to avoid copying if you can. The question is how much do you want to caller to understand about the lower layers.
1:29:30
fittestbits__
Well, if FAT, we're currently reading the whole file, so not fixed size blocks. But, we can change the stream code to use a fixed size buffer that always has the correct alignment.
1:31:51
fittestbits__
Well, it would be the block size that read and write are called with. Or, you could have the file system mount code set the cache block size as part of init.