9.2 scsi install: no fstab found
The Yast installer seems able to detect my hardware, and load the applicable modules, in particular the sym53c8xx SCSI driver for my adapter. All my disk partitions from my two disks are listed, but their types are all called "unknown". Selecting my root partition, sda7, manually doesn't help. Console 4 has lots of errors about SCSI parity, bus resets, etc., as though the driver is mishandling the adapter somehow. There are lots of other interesting (maybe) messages on some of the other consoles, but I can't think of any way to capture them. FWIW, when I first booted from the 9.2 CD #1, a box popped up asking me to insert modules disk 2, to which I clicked "back" in lieu of hitting the reset button (not having made a modules disk 2). The installation seemed to proceed from there, to the point where it could tell me the things mentioned above. When I finally told the graphical installer to abort, it put me back into linuxrc, so I checked what modules it had loaded. The sym53c8xx module was there, as well as several other SCSI modules. I removed the SCSI CD module, since my CD is EIDE, and things like Reiserfs and ext3 which I don't use. There remained a "mid-level SCSI" and a "disk driver" module, which I assume are needed by the sym53c8xx driver, but I'm not sure. Does anyone have any suggestions what to try next? Does this SCSI driver need some parameters in the 2.6 kernel? If so, how would I give them to the installer? I assume I don't really need to make a fleet of floppies to do the install, since I can boot from the CD drive, and in the thrashing I did, I seemed to get a list of installed modules. Many TIA, Jim
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hi Jim, <whack> | FWIW, when I first booted from the 9.2 CD #1, a box popped up | asking me to insert modules disk 2, to which I clicked "back" | in lieu of hitting the reset button (not having made a modules | disk 2). <whack> For some reason, your system is booting from one of the floppy images it found on the CD. This is ok -- that's why they're there -- but to proceed successfully you'll need to create and use the modules diskette's, too. The alternative to this feature might have left you unable to boot from the CD at all and, IMHO, it doesn't actually turn into a "fleet" until you have to make the three /boot/ floppies, too. ;-) Good luck & regards, - - Carl - -- ____________________________________________________________________ C. E. Hartung Business Development & Support Services http://www.cehartung.com/ carlh@cehartung.com Dover Foxcroft, Maine, USA Public Key #0x68396713 Reg. Linux User #350527 http://counter.li.org/ -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.5 (GNU/Linux) iD8DBQFCBPVmusxgymg5ZxMRAv5yAJwIzII+dUPRU415NxrwYzPUYvZBewCeJjun rhtR0Onyof2h2bTqFfAI0Uk= =Lw2G -----END PGP SIGNATURE-----
On Sat, Feb 05, 2005 at 11:33:43AM -0500, Carl E. Hartung wrote:
For some reason, your system is booting from one of the floppy images it found on the CD. This is ok -- that's why they're there -- but to proceed successfully you'll need to create and use the modules diskette's, too.
I created a diskette for Modules5, which has the sym53c8xx module, and selected Manual Installation, then told linuxrc to load that module. I was offered the opportunity to enter parameters for "scsi_transport_spi" as well as "sym53c8xx" but since I've never given any parameters to my boot process, I don't know what might be useful. Apparently the kernel is still not making contact with the SCSI subsystem; no sda or sdb shows up in /proc/mounts, the installation target still fails with "no fstab found," and the kernel messages show SCSI errors. I copied by hand this much from VC #4: Attached scsi generic sg1 at scsi0, channel 0, id 15, lun 0, type 0 sym0: scsi parity error detected: SCR1=132 DBC=50000000 SBCL=0 sym0:15: ERROR (81:0) (8-0-0) (1f/9f/0) @ (scripta 50:f31c0004). sym0: script cmd = 90080000 sym0: regdump: da 00 00 9f 47 ... ff ff ff sym0: SCSI BUS reset detected sym0: SCSI BUS has been reset Here's what the current SuSE 7.3/kernel 2.4.23 boot.msg shows: <6>sym53c8xx: at PCI bus 0, device 13, function 0 <4>sym53c8xx: setting PCI_COMMAND_PARITY...(fix-up) <6>sym53c8xx: 53c895 detected with Symbios NVRAM <6>sym53c895-0: rev 0x1 on pci bus 0 device 13 function 0 irq 11 <6>sym53c895-0: Symbios format NVRAM, ID 7, Fast-40, Parity Checking <6>sym53c895-0: on-chip RAM at 0xf4201000 <6>sym53c895-0: restart (scsi reset). <4>sym53c895-0: Downloading SCSI SCRIPTS. <6>scsi0 : sym53c8xx-1.7.3c-20010512 <4> Vendor: QUANTUM Model: QM39100TD-SW Rev: N1B0 <4> Type: Direct-Access ANSI SCSI revision: 02 <4> Vendor: IBM Model: IC35L018UCPR15-0 Rev: S80D <4> Type: Direct-Access ANSI SCSI revision: 03 <6>sym53c895-0-<0,0>: tagged command queue depth set to 4 <6>sym53c895-0-<15,0>: tagged command queue depth set to 4 <4>Attached scsi disk sda at scsi0, channel 0, id 0, lun 0 <4>Attached scsi disk sdb at scsi0, channel 0, id 15, lun 0 <6>sym53c895-0-<0,*>: FAST-40 WIDE SCSI 80.0 MB/s (25.0 ns, offset 31) <4>SCSI device sda: 17783249 512-byte hdwr sectors (9105 MB) <6>Partition check: <6> sda: sda1 sda2 sda3 sda4 < sda5 sda6 sda7 > <6>sym53c895-0-<15,*>: FAST-40 WIDE SCSI 80.0 MB/s (25.0 ns, offset 31) <4>SCSI device sdb: 35592156 512-byte hdwr sectors (18223 MB) <6> sdb: sdb1 sdb2 sdb3 sdb4 < sdb5 sdb6 sdb7 > I assume the fact that 9.2 says "sym0" instead of "sym53c895-0" and "scsi generic sg1 at scsi0" instead of "scsi disk sda at scsi0" is a sign of something really wrong, but I have no idea what to do to try to fix it. Does anything in the 7.3 boot.msg give a hint? The sym53c8xx module appeared in the list linuxrc offered me, and when I asked linuxrc to "show loaded modules," there it was. Fwiw, my /etc/fstab is actually a symlink to /etc/fstab.3, as a result of the multi-step process involved in adding sdb a while back. If you think Yast could be choking on that symlink, I could restructure it, but given the driver errors, I suspect my problem is not the symlink. I'd really appreciate any suggestions... Jim
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Jim Osborn wrote: <snip> | I created a diskette for Modules5, which has the sym53c8xx module, | and selected Manual Installation, then told linuxrc to load that | module. I was offered the opportunity to enter parameters for | "scsi_transport_spi" as well as "sym53c8xx" but since I've never | given any parameters to my boot process, I don't know what might | be useful. | | Apparently the kernel is still not making contact with the SCSI | subsystem; no sda or sdb shows up in /proc/mounts, the installation | target still fails with "no fstab found," and the kernel messages | show SCSI errors. I copied by hand this much from VC #4: <snip> Hi Jim, I'm pretty sure the symlink has nothing to do with your problem. In fact, a quick google on "sym53c8xx" pulls up bug reports from over the past year and before on this exact problem -- earlier kernels working, later ones not. I didn't dig deep enough to discern if or how this is solved, but the solution is probably out there by now -- you just have to dig deep enough. HTH & regards, - - Carl - -- ____________________________________________________________________ C. E. Hartung Business Development & Support Services http://www.cehartung.com/ carlh@cehartung.com Dover Foxcroft, Maine, USA Public Key #0x68396713 Reg. Linux User #350527 http://counter.li.org/ -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.5 (GNU/Linux) iD8DBQFCBiUdusxgymg5ZxMRAh03AJwMXizWc813jR/qIV6oBB5j1SOv5gCfcyhU asAZ+aM8LHbH0ovcD1QQR/4= =pRiK -----END PGP SIGNATURE-----
On Sun, Feb 06, 2005 at 9:09:34AM -0500, Carl E. Hartung wrote:
a quick google on "sym53c8xx" pulls up bug reports from over the past year and before on this exact problem -- earlier kernels working, later ones not.
I didn't dig deep enough to discern if or how this is solved, but the solution is probably out there by now -- you just have to dig deep enough.
Thanks, Carl. I'll proceed with that in mind; it hadn't occurred to me that the basics on the installation CD might be bad. A quick strategic question occurs, though: I'll need to get the corrected sym53c8xx module into the installation process, presumably by making a modules diskette in a format that linuxrc can recognize. The images on the CD seem to have a boot sector and a disk full of stuff that's not individual files. Will linuxrc take a module from a normal ext2 diskette? The happiest solution, of course, would be if SuSE has posted a corrected Modules5 image ready to copy, and I'll check that next. But on the chance they haven't (and I haven't seen mention of it on this list), am I SOL? I'd be surprised if my old gcc-2.95.3 will make the current kernel modules correctly, but maybe it will. How would I know? Could I count on compiler error messages to tell me? One of my primary motivations for installing 9.2 was to upgrade the compiler, and its libraries. It's frustrating to not even get the installer to talk to the hard disk. Might I simply have to give up on SuSE 9.2 and wait for (and hope for better luck with) 9.3? Thanks for all your help, Jim
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hi again, Jim Jim Osborn wrote: <snip> | Thanks, Carl. I'll proceed with that in mind; it hadn't occurred to | me that the basics on the installation CD might be bad. <snip> The CD is fine. You have two dove-tailing problems. In most cases they would have appeared -- and been solved -- separately; either at different times or on separate machines. Unfortunately, with your unique combination of hardware, you've run into both at the same time. Think of it this way: 1. If you were installing 9.2 on IDE instead of SCSI, you'd have made the module diskettes, used them to complete your installation and */maybe/* you'd be thinking about why your CD drive isn't letting you boot normally off the Linux CD. 2. If you were adding this SCSI controller into an already functioning 9.2 system, you'd /still/ be forced to deal with whatever the firmware / module / config issues are. So, try not to confuse one problem with the other. And please don't become prematurely discouraged or disappointed with SuSE *or* Novell *or* 9.2. This is a great distribution -- you just have a couple of unique 'humps' to overcome. ... A quick | strategic question occurs, though: I'll need to get the corrected | sym53c8xx module into the installation process, presumably by making | a modules diskette in a format that linuxrc can recognize. The | images on the CD seem to have a boot sector and a disk full of stuff | that's not individual files. Will linuxrc take a module from a | normal ext2 diskette? <whack> It just so happens that I've always used rawwrite.exe under Windows (Gasp!) to create the floppies. Why? Because the only time I need a Linux diskette is when the system that I need it for is down! The rest of the time stuff goes in and out of the box via TCP/IP, CDRW, etc,. Besides, MS makes a /great/ diskette duplication platform! ;-) Anyway, the first step here is to isolate the exact cause of your SCSI adapter problem. You may have to update the controller bios... or pass some parameters to the module before it loads... or, or, or. It's all conjecture at this point. If you'd like, get some of this research done and holler back. good luck & regards, - - Carl - -- ____________________________________________________________________ C. E. Hartung Business Development & Support Services http://www.cehartung.com/ carlh@cehartung.com Dover Foxcroft, Maine, USA Public Key #0x68396713 Reg. Linux User #350527 http://counter.li.org/ -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.5 (GNU/Linux) iD8DBQFCBpEJusxgymg5ZxMRAi1AAJ4lGo8kZHFNMlmx0+9ZCZ3TNyjpSgCfY10A dS6h6Ef5DsmSMP8OuJR5C9s= =a/B5 -----END PGP SIGNATURE-----
Carl E. Hartung a écrit :
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Hi again, Jim
Jim Osborn wrote: <snip> | Thanks, Carl. I'll proceed with that in mind; it hadn't occurred to | me that the basics on the installation CD might be bad. <snip>
The CD is fine. You have two dove-tailing problems. In most cases they would have appeared -- and been solved -- separately; either at different times or on separate machines. Unfortunately, with your unique combination of hardware, you've run into both at the same time.
Think of it this way:
1. If you were installing 9.2 on IDE instead of SCSI, you'd have made the module diskettes, used them to complete your installation and */maybe/* you'd be thinking about why your CD drive isn't letting you boot normally off the Linux CD.
2. If you were adding this SCSI controller into an already functioning 9.2 system, you'd /still/ be forced to deal with whatever the firmware / module / config issues are.
So, try not to confuse one problem with the other. And please don't become prematurely discouraged or disappointed with SuSE *or* Novell *or* 9.2. This is a great distribution -- you just have a couple of unique 'humps' to overcome. ... A quick | strategic question occurs, though: I'll need to get the corrected | sym53c8xx module into the installation process, presumably by making | a modules diskette in a format that linuxrc can recognize. The | images on the CD seem to have a boot sector and a disk full of stuff | that's not individual files. Will linuxrc take a module from a | normal ext2 diskette? <whack>
It just so happens that I've always used rawwrite.exe under Windows (Gasp!) to create the floppies. Why? Because the only time I need a Linux diskette is when the system that I need it for is down! The rest of the time stuff goes in and out of the box via TCP/IP, CDRW, etc,. Besides, MS makes a /great/ diskette duplication platform! ;-)
Anyway, the first step here is to isolate the exact cause of your SCSI adapter problem. You may have to update the controller bios... or pass some parameters to the module before it loads... or, or, or. It's all conjecture at this point. If you'd like, get some of this research done and holler back.
good luck & regards,
- - Carl
- -- ____________________________________________________________________ C. E. Hartung Business Development & Support Services http://www.cehartung.com/ carlh@cehartung.com Dover Foxcroft, Maine, USA Public Key #0x68396713 Reg. Linux User #350527 http://counter.li.org/ -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.5 (GNU/Linux)
iD8DBQFCBpEJusxgymg5ZxMRAi1AAJ4lGo8kZHFNMlmx0+9ZCZ3TNyjpSgCfY10A dS6h6Ef5DsmSMP8OuJR5C9s= =a/B5 -----END PGP SIGNATURE-----
Hi all, I just discovered this thread, I want to tell you, that I've a LsILogic 53C1010 scsi controller. SuSE 9.2 installed like a charm 2 months ago ! I was asked for the modules floppies. I stopped the install, created the floppies and restart the install. The sym53C8xx module is OK and really SuSE 9.2 is great. Michel.
On Sun, Feb 06, 2005 at 4:50:01PM -0500, Carl E. Hartung wrote:
The CD is fine.
Well, not if you're using a sym53c895 SCSI controller, apparently.
1. If you were installing 9.2 on IDE instead of SCSI, you'd have made the module diskettes, used them to complete your installation and */maybe/* you'd be thinking about why your CD drive isn't letting you boot normally off the Linux CD.
If I were using IDE instead of SCSI, I wouldn't be concerned with the sym53c895 SCSI module at all; the whole issue would be moot. Fwiw, I made the boot disk set and booted from floppy, just to eliminate the CD boot as a variable (all previous installs have been via boot diskette).
2. If you were adding this SCSI controller into an already functioning 9.2 system, you'd /still/ be forced to deal with whatever the firmware / module / config issues are.
Exactly. It seems SuSE 9.2 is contra-indicated for use with sym53c895 based systems. Perhaps, as another respondent indicated, the sym53c8xx module works with another SCSI controller that is not the sym53c895, but as it comes out of the box, SuSE 9.2 does not work with the sym53c895.
So, try not to confuse one problem with the other. And please don't become prematurely discouraged or disappointed with SuSE *or* Novell *or* 9.2. This is a great distribution -- you just have a couple of unique 'humps' to overcome.
Notice I didn't say "look for a different distro..." :) I'd like very much to be able to use it. I am disappointed that a fairly popular essential piece of hardware, that worked perfectly well under earlier SuSE distros has ceased to work in 9.2. I can see tripping up over this issue early on in the distro release process, with all the various bits of hardware to cover, but when the reports start coming in, I'd be a lot less disappointed if I saw evidence of a fix or a workaround emanating from SuSE. Connecting with the hard disk is a pretty fundamental part of installing a distro, a real show stopper.
| The images on the CD seem to have a boot sector and a disk full | of stuff that's not individual files. Will linuxrc take a module | from a normal ext2 diskette? <whack>
It just so happens that I've always used rawwrite.exe under Windows (Gasp!) to create the floppies. Why? Because the only time I need a Linux diskette is when the system that I need it for is down! The rest of the time stuff goes in and out of the box via TCP/IP, CDRW, etc,. Besides, MS makes a /great/ diskette duplication platform! ;-)
If you've got a spare MS box around. I'm all linux here. :) But how the diskette is copied is not the question; it's whether the image can be a simple module sym53c8xx.o, copied to a diskette with an ext2 filesystem on it and subsequently loaded into the installation process with linuxrc or whatever.
Anyway, the first step here is to isolate the exact cause of your SCSI adapter problem.
The most relevant threads I was able to find in English were several bug reports, and an 8-article discussion on alt.os.linux.suse that seemed to conclude that several posters had problems with the sym53c8xx module and the 2.6 kernel, whereas all had been well with the 2.4 kernels. No one had any workarounds to suggest. searching: site:suse.com sym53c8xx "9.2" gives 4 results, all in German. Some other search recipes give other German results, and one in Russian, if I'm not mistaken. Unfortunately, my language skills extend only to English.
You may have to update the controller bios...
Really? Would a newer kernel or driver require a bios upgrade? Are there even Linux tools to do that? That seems like a very hostile feature, that most driver maintainers would take pains to avoid.
or pass some parameters to the module before it loads... or, or, or. It's all conjecture at this point. If you'd like, get some of this research done and holler back.
Some posters referred to above said they'd tried various parameters, all to no avail. They didn't say just what parameters they'd tried. One poster mentioned the prospect of different interrupt handling in the 2.6 kernel being a potential problem. There was considerable discussion of various patching strategies for the sym53c8xx for use in a pcmcia environment. There may well be a workaround, possibly involving compiling a patched kernel, or a patched sym53c8xx module. If that's the case, I'd think SuSE might want to put up the results for offer on their site. The simplest workaround would be to offer installers the choice of falling back to the 2.4 kernel series, which seems to work splendidly. If moving to the 2.6 series renders the distro un-installable, maybe the 2.4 kernel would be the way to go. I can't think of anything else to research. If anyone from SuSE is following this thread, it'd be nice to know where someone in my position stands. I'd hate to waste a couple hundred hours chasing a dead end, which is what 9.2 looks like at this point. Does anyone have any other suggestions? One more procedural question: since the kernel messages are obviously stored somewhere by linuxrc and yast, it seems I should be able to copy them to a floppy before abandoning the installation. Does anyone know where they're stored? Has anyone saved those messages this way? If I persue this on the newsgroups, it'd be nice to be able to share the error messages I see; they're way too extensive, in way too tiny type for me to copy by hand. Jim
On Sun, Feb 06, 2005 at 4:50:01PM -0500, Carl E. Hartung wrote:
the first step here is to isolate the exact cause of your SCSI adapter problem. You may have to update the controller bios... or pass some parameters to the module before it loads... or, or, or. It's all conjecture at this point. If you'd like, get some of this research done and holler back.
Having installed the kernel 2.6.8 sources and docs from the 9.2 CDs, I've tried giving parameters to the sym53c8xx module during the 9.2 installation when linuxrc asks for such parameters. Specifically, I've tried to give these: mpar:n spar:n verb:2 debug:0x200 as discussed in section 10.2 of /usr/src/linux-2.6.8-24/Documentation/scsi/sym53c8xx_2.txt. I've tried the form above, as well as mpar=n spar=n... and both forms give the message: sym53c8xx: Unknown parameter `mpar' in /var/log/boot.msg. How can I give the scsi driver some parameters? TIA, Jim
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Jim Osborn wrote:
How can I give the scsi driver some parameters?
Hi Jim, Just a note to check in and let you know I haven't forgotten about this thread. I just haven't had a lot of time to sit down and go through the documentation. I did notice this subtle warning near the start: 4. Memory mapped I/O versus normal I/O Memory mapped I/O has less latency than normal I/O and is the recommended way for doing IO with PCI devices. Memory mapped I/O seems to work fine on most hardware configurations, but some poorly designed chipsets may break this feature. A configuration option is provided for normal I/O to be used but the driver defaults to MMIO. Have you investigated this and what have you tried, if anything, to eliminate it as a possibility? - - Carl - -- ____________________________________________________________________ C. E. Hartung Business Development & Support Services http://www.cehartung.com/ carlh@cehartung.com Dover Foxcroft, Maine, USA Public Key #0x68396713 Reg. Linux User #350527 http://counter.li.org/ -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.5 (GNU/Linux) iD8DBQFCFWQ6usxgymg5ZxMRAoMfAJ9xUv5/Ikw7fCdxXUE2VuiBGt03fgCdGdn6 V7c49PnEd8NbxCM4H+qmY4I= =ot9q -----END PGP SIGNATURE-----
On Thu, Feb 17, 2005 at 10:42:50PM -0500, Carl E. Hartung wrote:
I did notice this subtle warning near the start:
4. Memory mapped I/O versus normal I/O
Memory mapped I/O has less latency than normal I/O and is the recommended way for doing IO with PCI devices. Memory mapped I/O seems to work fine on most hardware configurations, but some poorly designed chipsets may break this feature. A configuration option is provided for normal I/O to be used but the driver defaults to MMIO.
Have you investigated this and what have you tried, if anything, to eliminate it as a possibility?
Thanks for the response, Carl. I haven't tried anything specifically for that option, but just tried some simple options that seemed like good starting points; since my problem seemed to report parity errors, I tried to turn off parity checking and see what happened; I also tried to turn on verbose reporting, that sort of thing. So far, I don't seem to be able to enter ANY parameters of ANY sort in linuxrc for this driver. Even just a simple "verb:2" as the one-and-only parameter shows: <3>sym53c8xx: Unknown parameter `verb:2' in /var/log/boot.msg. It's as though the driver has been compiled to recognize no options whatsoever. Very strange. I can't find any useful-sounding parameters in the scsi_transport_spi module sources or the scsi mid/low level driver doc, but that doesn't mean there aren't any, just that I can't seem to find them.
From the config.gz file that appears in /proc after yast is up and running, I find these lines, and only these, pertaining to (SYM|NCR)53C8:
# # SCSI Transport Attributes # CONFIG_SCSI_SPI_ATTRS=m CONFIG_SCSI_FC_ATTRS=m ... CONFIG_SCSI_SYM53C8XX_2=m CONFIG_SCSI_SYM53C8XX_DMA_ADDRESSING_MODE=1 CONFIG_SCSI_SYM53C8XX_DEFAULT_TAGS=16 CONFIG_SCSI_SYM53C8XX_MAX_TAGS=64 # CONFIG_SCSI_SYM53C8XX_IOMAPPED is not set so apparently SuSE 9.2 has only the sym53c8xx_2 driver, as a module, not the sym53c8xx (without the _2) driver. For comparison, my 2.4.23 kernel has this config: # # SCSI low-level drivers # # CONFIG_SCSI_SYM53C8XX_2 is not set # CONFIG_SCSI_NCR53C8XX is not set CONFIG_SCSI_SYM53C8XX=y CONFIG_SCSI_NCR53C8XX_DEFAULT_TAGS=4 CONFIG_SCSI_NCR53C8XX_MAX_TAGS=32 CONFIG_SCSI_NCR53C8XX_SYNC=20 CONFIG_SCSI_NCR53C8XX_PROFILE=y # CONFIG_SCSI_NCR53C8XX_IOMAPPED is not set # CONFIG_SCSI_NCR53C8XX_PQS_PDS is not set so at least the "_IOMAPPED is not set" parts agree. As you can see, I'm using the sym53c8xx driver, but from reading the changelogs, the sym53c8xx_2 driver should have all the capabilities of the sym53c8xx driver, unless SuSE has done something peculiar with it. FWIW, boot logs and dmesg outputs for a typical 9.2 attempt are at: ftp://ftp.eskimo.com/u/j/jimo/SuSE-9.2/2.6.8-24-logs and for comparison, the same info for my current 2.4.23 system: ftp://ftp.eskimo.com/u/j/jimo/SuSE-9.2/2.4.23-logs I'm booting from the 3-diskette set, and inserting modules diskette 5 for the sym53c8xx driver. The floppy0 errors don't appear until the modules disk, and occur with two different disks, neither of which shows errors during formatting. The diskette images were copied from CD with dd. I have not tried using insmod, or using any of the control commands via echo "<verb> <param>" >/proc/scsi/sym53c8xx/0 as the parity control seemed to be a boot-only option. I should play with those methods, I suppose, for lack of any other ideas. I'll be happy to provide any additional info you think might be interesting. Thanks for any help you can provide. Jim
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Jim Osborn wrote: <one big whack> Ha! Now _that's_ how you do your homework and bring somebody else up to speed! :-D I used to work at a RISC/*nix SCSI mass storage products manufacturer. You'd have felt right at home in engineering! Let me digest a bit (I've downloaded the logs) and I'll get back with you. regards, - - Carl - -- ____________________________________________________________________ C. E. Hartung Business Development & Support Services http://www.cehartung.com/ carlh@cehartung.com Dover Foxcroft, Maine, USA Public Key #0x68396713 Reg. Linux User #350527 http://counter.li.org/ -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.5 (GNU/Linux) iD8DBQFCFYLJusxgymg5ZxMRAvyjAJ4gfHKmgdVAh6qOrW+fhEr+AztOiwCfe491 Yw8LG4HXqCESaONKXl8GhTU= =kHeS -----END PGP SIGNATURE-----
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Jim Osborn wrote: <snip>
How can I give the scsi driver some parameters?
Hi Jim, I've looked at the logs and scanned the driver docs and think I may have one or two reasonable ideas. First, I'd like to locate and download a product data sheet and all the available documentation for that specific controller. I need the model number and a production date or rev. number -- and a URL, too, if you happen to know what it is. The driver's author refers to Symbios and Tekram setup programs for models where NVRAM is used to store the controller setup data. From sym53c8xx_2.txt, section 5: "If your controller has NVRAM, you can configure this feature per target from the user setup tool. The Tekram Setup program allows to tune the maximum number of queued commands up to 32. The Symbios Setup only allows to enable or disable this feature." There are other phrases scattered throughout the document which /seem/ to suggest that passing boot parameters isn't possible when the setup information in stored in NVRAM. Did you get a CD with that controller? If so, does it have a setup program on it? - - Carl - -- ____________________________________________________________________ C. E. Hartung Business Development & Support Services http://www.cehartung.com/ carlh@cehartung.com Dover Foxcroft, Maine, USA Public Key #0x68396713 Reg. Linux User #350527 http://counter.li.org/ -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.5 (GNU/Linux) iD8DBQFCFo+vusxgymg5ZxMRAtgCAJ4oqkZCLL0/Vj/bq7HBoeEBqv6D/ACfTa0y YvDnnRQ/M0qN9lSFnqUFB3Y= =Z7A2 -----END PGP SIGNATURE-----
On Fri, Feb 18, 2005 at 8:00:31PM -0500, Carl E. Hartung wrote:
First, I'd like to locate and download a product data sheet and all the available documentation for that specific controller. I need the model number and a production date or rev. number -- and a URL, too, if you happen to know what it is.
Hi Carl, According to the copy of the paper Users Guide VA Research gave me, it's a: Symbios SYM8951U Ultra2 SCSI Adapter. The revision record says: date: 01/97 remarks: rev 1.0 first official release which may refer more to the manual than to the board itself. I also seem to have a copy of the Symbios Logic PCI SDMS 4.0 User's Guide which I'll confess I haven't studied. It may have something interesting; at least it has sections entitled Device Driver for SCO Unix, and UnixWare. I'll dive into it and see what it says. There's a URL in that manual: http://www.symbios.com I don't know if the numbers in that boot.msg imply anything useful. The relevant section of my current 2.4.23 /proc/pci says: Bus 0, device 13, function 0: SCSI storage controller: LSI Logic / Symbios Logic 53c895 (rev 1). IRQ 11. Master Capable. Latency=247. Min Gnt=30.Max Lat=64. I/O at 0x1000 [0x10ff]. Non-prefetchable 32 bit memory at 0xf4202000 [0xf42020ff]. Non-prefetchable 32 bit memory at 0xf4201000 [0xf4201fff]. The system was purchased in Sept 1998. That's about all I have without cracking the case and extracting the card, which I'd rather not do if at all possible.
Did you get a CD with that controller?
No. They gave me CDs for just about everything else, but not the SCSI adapter, for some reason. Wish I'd thought to squawk at the time... Thanks again, Jim
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hi Jim, Here are some notes from my research and analysis that you should start reviewing: - --> Flashing the controller BIOS. The string "LSI Logic / Symbios Logic 53c895 (rev 1)" in your boot log is coming straight from the board. LSI Logic was formerly called Symbios Logic and this board was produced under both names. From the website: "... The BIOS and the flash utility are located at the LSI Logic web site in the Download Center under Support." I'm not sure this will fix your install problems, but '98 to now is quite a long time in computer years. And this definitely is a "rev 1" board. My assessment would be to use the most mature and field-tested code, which requires flashing the BIOS. Link to the page where all the links below came from: http://lsionline2.lsil.com/esupport/esupportlsi/consumer/esupport.asp? id=38e2a3ae-c38e-4fb1-b682-90eccd37e9a6&resource=&number=2&isExternal= 0&nShowFacts=&nShowCause=&nShowChange=&nShowAddInfo=&activepage= statement.asp&bForceMatch=False&strCurrentSymptom=&searchtype= normal&searchclass=QuickSearch&bnewsession=false&selecttype=match (Chopped at "?" and "=" to avoid random word-wrap breaks) - --> Link to Symtools utility for SDMS: http://www.lsilogic.com/downloads/license.do?id=2000&did=5204&pid=4112 - --> Link to Symtools utility for SDMS readme document: http://www.lsilogic.com/files/support/ssp/sdms/Utilities/SCSITOOL.TXT - --> Link to their Linux driver version 1.7.3a from 09/09/02: http://www.lsilogic.com/downloads/license.do?id=2000&did=2916&pid=4112 - --> Link to the Linux driver readme document: http://www.lsilogic.com/files/support/ssp/sdms/linux/linux8xx.txt regards, - - Carl - -- ____________________________________________________________________ C. E. Hartung Business Development & Support Services http://www.cehartung.com/ carlh@cehartung.com Dover Foxcroft, Maine, USA Public Key #0x68396713 Reg. Linux User #350527 http://counter.li.org/ -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.5 (GNU/Linux) iD8DBQFCFtFKusxgymg5ZxMRAk7wAJ9X5TvNaBrllL3pwU/NR7lXfIyBjwCdFpTb 2yXZelzyZrdTV/pv4deA73g= =u+Ph -----END PGP SIGNATURE-----
On Sat, Feb 19, 2005 at 12:40:27AM -0500, Carl E. Hartung wrote:
I'm not sure this will fix your install problems, but '98 to now is quite a long time in computer years. And this definitely is a "rev 1" board. My assessment would be to use the most mature and field-tested code, which requires flashing the BIOS.
Thanks, Carl, I'll check those links and grab the utilities. Some questions remain, though: First, wouldn't I be risking losing the machine (and with it my ability to connect and call for help) altogether if the new BIOS doesn't allow a full 9.2 install, and at the same time, breaks my ability to boot back to my current installation? Is the flash utility likely to include the ability to Restore-Exactly-What- Was-There-Previously? If it does, then it's no real concern, but I'd hate to have to order a new SCSI board to fix a failed experiment, particularly without a computer to help with the order. I live a couple of hours from the big city... Second, while the current rev 1 board may be old and creaky, it works well with the kernel 2.4.23 sym53c8xx driver. Obviously a newer BIOS could require a newer driver, and I can understand that a newer BIOS could provide improved performance, but I can't understand how a newer Linux driver could fail altogether to operate the older board. What am I missing in that thought process? The sym53c8xx_2 changelog ends with 2.1.17a, and the boot.msg says 2.1.18j, so maybe there's something in there about invalidating older BIOSes, but I sure don't see anything like that up to 17a. My current sym53c8xx says 1.7.3c in the boot.msg, which is the most recent in the SuSE 9.2 sym53c8xx changelog. One of the entries in that bit of the c-log says: - Backport from SYM-2 the work-around that allows to support hardwares that fail PCI parity checking. as though my sym53c8xx and the new sym53c8xx_2 drivers should both know how to deal with a problem suspiciously like mine. Finally, while the BIOS issues are interesting and may prove useful, isn't the failure of the parameter(s) to the SCSI module occurring before the driver knows anything about the BIOS? Am I not dealing with a linuxrc/SuSE-kernel issue there, something to do with the format of the options/parameters, a mismatch between the docs and the actual driver code, or something at that level? It seems I should be able to turn on verbose reporting, at least, regardless of the age of the card BIOS. Doesn't that extra information come from the driver, not the card? The "Unknown parameter" error in the boot log occurs before the <6>sym0: <895> rev 0x1 at pci 0000:00:0d.0 irq 11 <4>sym0: Symbios NVRAM, ID 7, Fast-40, LVD, parity checking lines, which would seem to indicate first contact with the board itself. I don't mean to sound like I really know what I'm talking about here, but these questions seem relevant and if you could straighten me out on them, it'd really help me understand what I'm dealing with. In the meantime, I'll follow those links, and maybe take another pass through the SuSE sym53c8xx_2 sources. Thanks for the help. Jim
Is the flash utility likely to include the ability to Restore-Exactly-What- Was-There-Previously?
Yes, it should ask if you want to backup your existing bios into a .bin file of some kind, if it all goes t-up, as long as you can boot up onto your flash disk you can restore the bios. As for if it doesn't post, old boards usually had a failsafe in them that they would boot in a "safemode" of sorts after three failed post attempts, designed for people who accidently fried their bios'es by flashing the wrong one. But if your careful you shouldn't need any of that :)
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hi Jim, I posted the Linux driver and "Symbios tools utility" links but forgot the BIOS flash utility. - -->Link to the flash program: http://www.lsilogic.com/downloads/license.do?id=2000&did=2839&pid=2234 - --> Link to the flash program readme doc: http://www.lsilogic.com/files/support/ssp/sdms/Bios/FLASH.TXT Upon further review, the Symbios tools utility was pretty much window dressing when it was released. It hasn't gotten any better with age, either. I wouldn't bother with it unless you expect to be running Win95 or NT 1.10.00 <g> - - Carl - -- ____________________________________________________________________ C. E. Hartung Business Development & Support Services http://www.cehartung.com/ carlh@cehartung.com Dover Foxcroft, Maine, USA Public Key #0x68396713 Reg. Linux User #350527 http://counter.li.org/ -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.5 (GNU/Linux) iD8DBQFCFtkFusxgymg5ZxMRAkHFAJ90qKZFtLvJ+uvPKl5LR0s/RnOn2gCeKnwB 4sL7bNZX1Tsnja+9NfXnafQ= =Rw8v -----END PGP SIGNATURE-----
On Sat, Feb 19, 2005 at 1:13:25AM -0500, Carl E. Hartung wrote:
I posted the Linux driver and "Symbios tools utility" links but forgot the BIOS flash utility.
- -->Link to the flash program: http://www.lsilogic.com/downloads/license.do?id=2000&did=2839&pid=2234
- --> Link to the flash program readme doc: http://www.lsilogic.com/files/support/ssp/sdms/Bios/FLASH.TXT
A quick glance at that doc says that it does include the ability to save and restore the existing BIOS; that's the good news. The bad news is the opening line: "This document provides information for the PCI Flash Utility V1.14.00 for DOS." and looking through it, they seem to really mean it: DOS, no Linux anywhere in sight. I see only that one flash download on their page, and the zip file contains only the one .exe file, nothing Linuxy. This machine doesn't do DOS, so it seems that flash program won't help; let's hope I can get SuSE 9.2 to install with the current BIOS. Cheers, Jim
Jim Osborn wrote:
On Sat, Feb 19, 2005 at 1:13:25AM -0500, Carl E. Hartung wrote:
I posted the Linux driver and "Symbios tools utility" links but forgot the BIOS flash utility.
- -->Link to the flash program: http://www.lsilogic.com/downloads/license.do?id=2000&did=2839&pid=2234
- --> Link to the flash program readme doc: http://www.lsilogic.com/files/support/ssp/sdms/Bios/FLASH.TXT
A quick glance at that doc says that it does include the ability to save and restore the existing BIOS; that's the good news. The bad news is the opening line:
"This document provides information for the PCI Flash Utility V1.14.00 for DOS."
and looking through it, they seem to really mean it: DOS, no Linux anywhere in sight. I see only that one flash download on their page, and the zip file contains only the one .exe file, nothing Linuxy.
This machine doesn't do DOS, so it seems that flash program won't help; let's hope I can get SuSE 9.2 to install with the current BIOS.
http://www.bootdisk.com/ if "doesn't do DOS" is simply because DOS is not currently installed anywhere on the system, or you don't already have some DOS boot floppy you can use long enough to run some flash program. If you can't boot a floppy, then you can temporarily move the HBA to some machine that can boot a DOS floppy. The preferred method of LSI flashing is with no cables connected to the HBA. All that said, I seriously doubt there's any liklihood that flashing will solve any of your problems. I may have had similar trouble. I say "may" because I don't remember exactly what my problem was when I couldn't install SuSE 9.1. All my earlier SuSE installs to SCSI were with 2.4 kernels. My only successful 2.6 kernel installs to SCSI have been on one system used for Mandrake Cooker that uses a slow old SYM8751SP non LVD HBA. I have 2 SuSE 9.2 systems now. Both require the sym53c8xx driver for SCSI HD access. This machine is one of them, and uses the LSI21002 dual channel HBA. 9.2 is installed to hda, in addition to OS/2 and PC DOS. 9.2 takes as long as 32 minutes to boot 9.2, and has SCSI bus error delays using various utilities, if I have a certain Quantum Fireball HD on the 8 bit bus. Mandrake 10.2 has similar problems with a newer 2.6.10 kernel, though not as bad. Removing the Quantum removes all the trouble that I am aware of. DOS, Knoppix (2.4 kernel) and OS/2 have no SCSI problems. My other 9.2 system also uses the LSI21002. I originally tried to install 9.1 to a Maxtor Atlas IV on it. Inexplicable SCSI installation snags, the details of which are long forgotten, led me to try scsitool.exe. The scsitool diagnostic told me there were bus problems and pointed me to the Maxtor. I called Maxtor about it. Maxtor replaced it with an Atlas V. The Atlas V improved nothing. So, I gave up and installed to hda instead of the Maxtor on sda. Once installed, the Maxtor seemed to give no problem to booted 9.1 and seems to give no problem to booted 9.2. What I do remember about the failed 9.1 install attempts was substituting other older SCSI HBAs eliminated the bus errors. I don't remember exactly which I substituted, but they could have been any of SYM8951, SYM8952, SYM8953; and also SYM8751. The 8751 was suitable for testing purposes only, as it provides suboptimal performance for the LVD HDs. The SYM895x devices also were suitable only for testing, because they have only the 68 pin bus, and the system uses SCSI for non-HD devices, including CD for installation. Scsitool gave the errors only with the 21002, and only with the Maxtor drives. Substiting an IBM HD eliminated the bus errors during scsitool tests. Oh, and I have all my LSI (SYM) HBAs flashed to the latest LSI BIOS. -- "He who believes in me will live, even though he dies; and whoever lives and believes in me will never die." John 11:25 NIV Team OS/2 ** Reg. Linux User #211409 Felix Miata *** http://members.ij.net/mrmazda/
On Sat, Feb 19, 2005 at 6:26:27AM -0500, Felix Miata wrote:
I seriously doubt there's any liklihood that flashing will solve any of your problems... All my earlier SuSE installs to SCSI were with 2.4 kernels.
Thanks for the response, Felix. Since my HBA and my two SCSI HDs work well with the 2.4.23 kernel and its sym53c8xx-1.7.3c driver, I think if I'm to get this machine working with SuSE 9.2, it won't need a new HBA BIOS. The question at this point is: can I figure out what it takes to make 9.2 install in less time than I can learn what it takes to upgrade a 2.4-based system to the latest C compiler, libraries, applications, etc., many of which I fear are going to be expecting a 2.6 kernel? And being on a dialup connection, I'd prefer not to spend weeks downloading all that stuff; I'd really prefer to have it all on a set of CDs compiled against the right libraries. Like SuSE 9.2. The reason I continue to beat my head against this installation is that it seems tantalizingly close. I keep thinking maybe if I can redirect one interrupt somewhere, or one dma channel, or figure out the parameter to give to the HBA, that'll make it all work. For example, what's with the flood of floppy0 errors when loading the modules from the diskette? Is the image on the CD bad? Does it require some special block size to dd? Is there some parameter I should be giving at initial boot time to deal with that? There are no complaints while loading the boot diskettes, just the modules, from both the modules2 and modules5 images. And what's with the "Unknown parameter" errors to the SCSI module? The fact that they occur before the announcement of the HBA make and model makes me think they're part of SuSE's linuxrc, not as a result of contact with the HBA itself. Do I need to give the parameters in French? If I could make any of the adjustments alluded to in the 2.6.8 docs on the SuSE CD, it seems I should be able to get this thing to work. Are the floppy0 errors corrupting the sym53c8xx_2 module just enough to make it refuse parameters and generate parity errors on one, and only one, of the HDs? This is my one-and-only machine, so I don't have the luxury of doing much hardware swapping, and since the HBA works so well currently, I hate to just order another brand on spec. Your experience makes me worry that even if I got a new motherboard and controller, I might be facing equally difficult issues. I think I'll put some effort into contacting the driver author, Gerard Roudier, in hopes of his telling me whether persuing his driver with the 2.6.8 kernel is a waste of time. The one other respondent in this thread besides Carl said they had been able to get 9.2 to install on a 53C1010 HBA, so apparently there's at least one Lsi HBA working. Maybe they'll report if they gave the module any parameters, and if so, what format they were given in... Cheers, and thanks again for the response, Jim
For some reason, your system is booting from one of the floppy images it found on the CD. This is ok -- that's why they're there -- but to proceed successfully you'll need to create and use the modules diskette's, too.
Sorry, Carl, but I think kernel modules are located on root disk image (/boot/root - 57 MB on SuSE 9.2 DVD). Modules 1, 2, etc only needed if one want to boot from floppies, not DVD. Please correct me if I am wrong. I have very similar problem - SuSE 9.2 do not recognizes SATA hardware RAID neither on Intel 875, neither on 7220 m/b. ********************************************* * Best Regards --- Andrei Verovski * * Personal Home Page * http://snow.prohosting.com/guru4mac/ * Mac, Linux, DTP, Development, IT WEB Site *********************************************
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Andrei Verovski (aka MacGuru) wrote: | |> For some reason, your system is booting from one of the floppy images it |> found on the CD. This is ok -- that's why they're there -- but to |> proceed successfully you'll need to create and use the modules |> diskette's, too. |> | | Sorry, Carl, but I think kernel modules are located on root disk image | (/boot/root - 57 MB on SuSE 9.2 DVD). Modules 1, 2, etc only needed if | one want to boot from floppies, not DVD. Please correct me if I am | wrong. I have very similar problem - SuSE 9.2 do not recognizes SATA | hardware RAID neither on Intel 875, neither on 7220 m/b. Hi Andrei, I appreciate your feedback, but Jim's (the OP's) system prompted him to insert modules diskettes after being booted from the install CD. As I understand it, this only occurs when the boot component of the primary image is refused or unrecognized by the system. The alternates are installed on the CD for just these situations. Which of the alternates was being picked up is unclear to me -- as I haven't studied this process in detail for 9.2 yet -- but the main focus of my post was to reassure Jim (the OP) and explain this fallback installation concept to him. Thank you for paying attention, though, 'cause the more input the better under these circumstances. regards, - - Carl - -- ____________________________________________________________________ C. E. Hartung Business Development & Support Services http://www.cehartung.com/ carlh@cehartung.com Dover Foxcroft, Maine, USA Public Key #0x68396713 Reg. Linux User #350527 http://counter.li.org/ -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.5 (GNU/Linux) iD8DBQFCB3dPusxgymg5ZxMRAlRVAJ40MgbNA9MzrYCWbvZwCy+TsY0USQCeNCE1 XbBVqasLs2DNpLci4i7Z7H4= =yqew -----END PGP SIGNATURE-----
participants (6)
-
Andrei Verovski -
Carl E. Hartung -
Catimimi -
Felix Miata -
Jim Osborn -
Stephen Furlong