Putting zhen on CC because he was the one who asked for SATA testing. I decided to stick my IDE hard disk on a IDE-->SATA converter in order to test out if the LiveCD's work with SATA or not. Here is my experience. I first tried the normal "gentoo" kernel. System booted up, and /dev/hde was created as my SATA drive. I mounted this on /mnt/gentoo and did some read/write tests, it seemed to be working OK except from my console being flooded with discouraging error messages on every operation. I'll upload the "dmesg" output that I could get. Sorry, I didn't get the whole lot because simply writing the file to disk exceeded the buffer. I next tried the "smp" kernel. Booting up was slow (6 minutes stuck on 1 part), as it was probing hdg (which presumably is my second SATA port with nothing connected). This issue has been solved in newer kernels, so as long as .2 uses 2.6.6 or newer then we should be OK here. Again I got some strange messages on the console, reading and writing worked ok. I saved the dmesg output and will upload this. Unfortunately my SATA experience in 2.6.6-mm2 wasn't much better even on my local install, I got an oops. I'll try with 2.6.6 next and see what happens.
Created attachment 31495 [details] dmesg output from "gentoo" kernel
Created attachment 31496 [details] dmesg output from "smp" kernel
OK, I have sorted it out. Those messages that I posted in the dmesg output are caused by the "siimage" SATA driver. This driver is buggy, unmaintained, and messages like the ones I posted are very common. The proper way to do it is to use Jeff Garzik's libata drivers (the ones under CONFIG_SCSI_SATA). I am writing this message from a SATA-booted system using the CONFIG_SCSI_SATA_SIL driver. So, all you have to do to get SATA working for me on the LiveCD's is compile the kernel with libata drivers and without siimage. The drives then identify as SCSI, e.g. /dev/sda (this was "hde" on the siimage driver, might be worth pointing out) FYI, my SATA controller is: 0000:01:0b.0 RAID bus controller: Silicon Image, Inc. (formerly CMD Technology Inc) Silicon Image Serial ATARaid Controller [ CMD/Sil 3112/3112A ] (rev 02) This type of controller is common on many nforce2 based motherboards.
*** Bug 40974 has been marked as a duplicate of this bug. ***
*** Bug 49485 has been marked as a duplicate of this bug. ***
This has been changed in the most recent (smp) kernel for the livecd's. We'll need testing shortly.
*** Bug 52450 has been marked as a duplicate of this bug. ***
I am adding amd64 to this also, since people are having problems with those systems too. (Probably the siimage sil_sata problem)
Would everybody test out the http://dev.gentoo.org/~wolf31o2/releng CD and let me know how it works for you with SATA in this bug... thanks
*** Bug 55033 has been marked as a duplicate of this bug. ***
The truth is that both drivers are buggy and incomplete. The bug filing that was just marked as duplicate was an observation that the siimage driver is entirely _not_included_ on the 2004.2-test4 CD. It should at least be included to allow trying of both drivers. I'm not saying that it should be automatically loaded, just included.
Are you sure you have the two the right way round? Through all my searching, I didn't find any users having "real" problems with sata_sil, whereas siimage was the total opposite. Does your hardware have trouble with sata_sil? If so, I'd suggest you send off a bug report to Jeff Garzik so that we can get it fixed, we can't easily do that for siimage as its been left to rot...
As for siimage not being on the CD, you're correct, and that was my fault. It will be on the -test5 CD, just not compiled statically in the kernel (only as a module).
i just tried out -test4 iso and it almost works for me with sil3112 hardware i just need a newer kernel 2.6.6 at least iirc latest -bk wouldnt hurt tho ;) when is -test5 going to be ready ?
Sorry that I didn't get back to this bug sooner. Have you tried any of the other -test series CDs? This should all be fixed in the 2004.2 release. Anyway, I'm going to mark this as RESOLVED-CANTFIX simply because I can't change the 2004.1 release, and we've finalized 2004.2 already.
