#84: grub2 can boot linux from IDE disk ---------------------------------+------------------------------------------ Reporter: oxygene | Owner: oxygene Type: enhancement | Status: new Priority: major | Milestone: Port GRUB2 to LinuxBIOS Component: code | Version: v3 Keywords: | Dependencies: Patchstatus: there is no patch | ---------------------------------+------------------------------------------ Grub2 uses BIOS calls right now for access to disk. Replace this with a driver that accesses ATA disks directly.
On Tue, Jul 10, 2007 at 07:37:34AM -0000, LinuxBIOS wrote:
Grub2 uses BIOS calls right now for access to disk. Replace this with a driver that accesses ATA disks directly.
_Please_ don't get the one from FILO.
Either grab the one from OpenBIOS or port either ADVDRVR or MINDRVR written by Hale Landis, they're both in the public domain and available from http://ata-atapi.com/
(How does this thing work to get trac to put emails as ticket comments? trac docs anywhere?)
//Peter
On Thu, Jul 12, 2007 at 07:53:33PM +0200, Peter Stuge wrote:
On Tue, Jul 10, 2007 at 07:37:34AM -0000, LinuxBIOS wrote:
Grub2 uses BIOS calls right now for access to disk. Replace this with a driver that accesses ATA disks directly.
_Please_ don't get the one from FILO.
Either grab the one from OpenBIOS or port either ADVDRVR or MINDRVR
Why not FILO's driver? Are those other two better in some respect? How/why?
I've seen lots of inline assembly in ADVDRVR upon a quick investigation.
written by Hale Landis, they're both in the public domain and available from http://ata-atapi.com/
(How does this thing work to get trac to put emails as ticket comments? trac docs anywhere?)
Doesn't work. It works the other way around, though. If you post comments to the trac issue, they'll be (automatically) posted on the list, too.
http://trac.edgewall.org/wiki/TracGuide
Uwe.
On Thu, Jul 12, 2007 at 09:06:28PM +0200, Uwe Hermann wrote:
Grub2 uses BIOS calls right now for access to disk. Replace this with a driver that accesses ATA disks directly.
_Please_ don't get the one from FILO.
Either grab the one from OpenBIOS or port either ADVDRVR or MINDRVR
Why not FILO's driver?
Because it's a somewhat unreliable patchwork with lots of hacks. (In a bad way.)
Are those other two better in some respect? How/why?
The OpenBIOS driver is very nice and neat, but not as featureful as ADVDRVR which even does DMA.
I've seen lots of inline assembly in ADVDRVR upon a quick investigation.
Aha. Maybe MINDRVR is a better fit. It would be interesting to see a feature comparison between the OpenBIOS driver and MINDRVR.
Thanks.
//Peter
Peter Stuge wrote:
The OpenBIOS driver is very nice and neat, but not as featureful as ADVDRVR which even does DMA.
I'm pretty much planning to use openbios' (axboe's) driver. I also doubt that DMA is useful in the boot environment, and I'd like to keep complexity down.
Regards, Patrick Georgi
* Uwe Hermann uwe@hermann-uwe.de [070712 21:06]:
(How does this thing work to get trac to put emails as ticket comments? trac docs anywhere?)
Doesn't work. It works the other way around, though. If you post comments to the trac issue, they'll be (automatically) posted on the list, too.
You might try sending to trac at linuxbios dot org. The email filter is there but it's long since we installed it, no idea whether it works.
Stefan
On Thu, Jul 12, 2007 at 10:28:30PM +0200, Stefan Reinauer wrote:
You might try sending to trac
But what should I send? Any docs for the filter?
//Peter
On Thu, Jul 12, 2007 at 10:19:30PM +0200, Patrick Georgi wrote:
Peter Stuge wrote:
The OpenBIOS driver is very nice and neat, but not as featureful as ADVDRVR which even does DMA.
I'm pretty much planning to use openbios' (axboe's) driver.
Are the grub2 fs drivers already written to do multiblock reads from the block devices btw? This is pretty important as seen in FILO. (Admittedly a design decision above the IDE driver.)
I also doubt that DMA is useful in the boot environment, and I'd like to keep complexity down.
I think DMA would be nice. It can make a difference already with a few megabytes which isn't very uncommon for a complete kernel and easily reached with a ramdisk or xen.
The complexity is indeed the tradeoff, but it has been suggested that IDE DMA need not be all that complex.
Perhaps worthwhile to have a look at what MINDRVR does? 975 loc
//Peter
#84: grub2 can boot linux from IDE disk ----------------------------+----------------------------------------------- Reporter: oxygene | Owner: oxygene Type: enhancement | Status: closed Priority: major | Milestone: Port GRUB2 to LinuxBIOS Component: code | Version: v3 Resolution: fixed | Keywords: Dependencies: | Patchstatus: patch needs review ----------------------------+----------------------------------------------- Changes (by oxygene):
* status: new => closed * patchstatus: there is no patch => patch needs review * resolution: => fixed
Comment:
contained in lbgrub2 patch, using code from openbios (unclear licensing, gpl2 vs. gpl3!)
see http://openbios.org/~oxygene/patch-20070820-2-lbgrub2-genesis-patch and http://openbios.org/~oxygene/lbgrub2-20070820-1.tar.bz2
-
LinuxBIOS -
Patrick Georgi -
Peter Stuge -
Stefan Reinauer -
Uwe Hermann