freenode/#mezzano - IRC Chatlog
Search
14:33:24
froggey
can you try booting with the freestanding option? that should produce a more useful set of backtraces
14:34:47
froggey
at the kboot boot menu there should be a "Mezzano - freestanding" option, pick that one
14:49:40
froggey
Shinmera: try going to the kboot shell (F10 at the menu) and booting with: mezzano "hd0,2" "freestanding" "no-detect"
14:50:39
Shinmera
It'd be nice if the menu lasted more than one second, clicking into the vm and inputting before it's gone is a bit tough.
14:54:42
froggey
you could try mounting the bootloader partition in the image and editing the boot config that way...
15:08:50
froggey
I guess you could dig through the ata spec and figure out exactly how to detect the sector size properly
15:12:25
froggey
actually there's no support for any of the vmware nics, so building world be impossible
15:14:02
Shinmera
If I didn't have too much on my plate (or enough helpers to manage it) I would be interested.
15:22:56
froggey
looks like vbox & vmware both support the Intel 82545EM, so that'll be the one to target. I'll do that... eventually
15:27:47
froggey
it's a shame the intel gpu on my test machine is so old. intel releases documentation, but it doesn't stretch quite that far back
15:29:51
froggey
I was able to puzzle out the 2D/modesetting side from the available docs, but the 3D side looks totally different. I'm not sure it even supports hardware vertex shaders...
15:47:21
random-nick
does AMD release documentation or do they just release the linux driver and firmware?
16:05:39
alexshendi
Which network cards are supported on real HW? I managed to run Demo 5 on a Thinkpad X220.
19:41:02
alexshendi
Hi, it's me, again. How much work would it be to write a driver for the intel gigabit Ethernet adapter in Mezzano?
19:44:00
froggey
but generally it's not too difficult getting a NIC working, at least with a relatively simple & slow driver
19:45:17
alexshendi
froggey, no, I haven't. My idea was to have a look at the openbsd driver, understand what it does and rewrite it in CL.
19:45:33
Shinmera
froggey: Recently I wrote a library for file attribute access (uid/gid/flags/atime/mtime/btime). How should I implement these functions for Mezzano?
19:48:53
froggey
alexshendi: probably not very comprable, but I haven't looked at any openbsd drivers so I don't really know
19:51:38
Shinmera
The best would be to have some kind of interface that you would use for all file systems you implement.
19:51:59
Shinmera
now, capabilities between FSs varies a lot, so I can't say right away what the best course would be.
19:52:10
froggey
ah, right. that's basically how file-write-date/etc work. they call generic functions based on the pathname
19:53:32
Shinmera
For the attributes I currently expose a 'raw integer' plus a standardised set of translation functions that interpret that integer.
19:54:37
Shinmera
For instance I'd be fine with just returning a static uid/gid if you don't intend on ever supporting multiuser stuff.
19:56:03
froggey
unix uid/gids aren't something I'd add specific functions for, maybe accessible by some more abstract file-property function
19:57:19
froggey
but file system stuff isn't something I really care a whole lot about in general...
19:57:40
froggey
fittestbits__ and ebrasca are the ones who've been doing the most on it, probably better to talk to them
19:58:38
Shinmera
alexshendi: You might want to look at the Barrelfish OS. They have an abstract device driver description language that might be useful. Dunno if they have a description of the one you want to implement, though: http://git.barrelfish.org/?p=barrelfish;a=tree;f=devices;h=6321a7f596d29f47c107ae94ba73a2d7ccc3854f;hb=HEAD
20:00:44
Shinmera
froggey: To be honest, I was mostly looking to implement file-attributes so I could see whether Zippy would run on Mezzano :)
20:15:31
Shinmera
The funnest part about device drivers is that often devices won't adhere to spec :)
20:22:35
alexshendi
I am wondering about USB Ethernet and WiFi drivers. Would they be easier to implement?
20:23:11
froggey
not sure about USB Ethernet, you'd need to talk to fittestbits__ about the status of the USB stack
20:23:55
froggey
you'd also need one of the supported USB controllers (OHCI or EHCI, no driver for UHCI)
20:27:11
fittestbits__
About the file attributes - CL provides a couple of discrete functions: file-write-date and file-author. But, adding a discrete function for all of the attributes seems awkward
20:30:36
fittestbits__
Sorry, I meant there's a generic driver for OHCI and EHCI USB controllers - but no specific drivers for Ethernet or WiFi.
20:31:25
fittestbits__
The generic drivers provide APIs for registering a device or class driver and for sending/receiving packets.
20:31:56
fittestbits__
There is a class driver for mass storage, and a HID class driver that supports mice.
20:34:57
fittestbits__
The mass storage driver is relatively simple - some code to parse configuration info from the device and then read and write routines.
20:36:34
fittestbits__
The HID code is complicated because the device configuration information contains a description of the messages that the device generates.
20:37:54
fittestbits__
alexshendi: where did I lose you ... happy to answer questions or explain more ...
20:38:49
froggey
a single function for file properties seems good, but multiple values would be a bit restrictive. returning a plist would let file system return custom attributes
20:40:32
fittestbits__
That makes sense. A plist has good flexibility - not like a structure or class or multiple values ...
20:42:19
alexshendi
I am sorry, I really don't know much about either Ethernet adapters or USB in general.
20:44:00
fittestbits__
OK - OHCI controllers support USB 1.1, EHCI controllers support USB 2 and often (always?) come with an OHCI controller for USB 1.1. and XHCI controllers support USB 3.
20:44:01
alexshendi
fittestbits__, froggey: Thank you for your explanations. I must go to bed now, amm really tired. Good night!
20:57:38
stylewarning
Is there a tracker somewhere for minimum drivers needed to run on Real Hardware?
21:00:02
froggey
my test machine is an Intel D945GCLF2 and there are drivers for everything but the UHCI controller
21:18:01
fittestbits__
froggey: merged up to the latest code and boot is failing - hanging at the end of the PCI scan. This is on VBox.
21:23:59
froggey
ok. I think the new spurious IRQ detection code is too eager, try without the second controller for now if you can
22:36:02
froggey
oh whoops. I had the sense of the PIC In Service Register inverted, it was treating non-spurious IRQ7/15s as being spurious and spurious ones as non-spurious
22:43:23
froggey
fittestbits__ and Shinmera, if you could take a look and let me know if that seems like a sensible design
22:44:00
froggey
fittestbits__: also I've added timeouts to the ATA & AHCI interrupt paths, so that IRQ sharing issue you were having on real hardware should be less of a problem