First of all I want to say thank you to Jens Axboe, Peter Osterlund, Ben Fennema, and all the others, who participated in making Linux packet writing avaiable. I first tried the patch long time ago without success (when it was maintained by Axboe). Now I had the time to look further into the issues and finally I managed to make it work. I'm using an Acer 4x4x32x IDE-ATAPI CDRW connected to the PIIX4 ide controller of my Abit BP6 motherboard (it also has a hpt366). Let me mention, that this writer shows some strange symtoms: firmware updater programs haven't been able to find the writer on the ide bus for some time and the writer does not care about the specified grabbing and writing speeds, it just operates at approx. 2,5x speed. The latter one doesn't refers to packet writing, when it seems to keep the desired speed. It can handle rewritable discs well in Windows (using InCD or DirectCD). Once (in the past: 2 year ago+) it created some disks with errors. I tried to stress it with writing 10 CDs to crash it totally before I lost the guarantee. It created dosens of CDs after that without errors. Now I continue with my recent experiences. I compiled the kernel with packet-2.4.19-2.patch. My kernel based on a 2.4.19-rc1 gentoo kernel with some additional patches from Andrea Arcangeli. Because it is an -ac series kernel from the packet point of view, I modified slightly the patch based on the information appeared on this mailing list. I'm using udftools-1.0.0b2. The writer resides on the PIIX4, because it is a more stable product than the hpt366 IMHO. I didn't tune the writer with hdparm. It didn't change anything. Modules were loaded OK. When I tried to quick-setup the disk with cdrwtools, everything seemed to be fine. It produced a CD which I couldn't mount. Some I/O errors showed up and it said that no UDF filesystem could be found. After some retries I tried to format the disk with mkudffs. At the first attempt it flooded the screen with read errors. When I repeated this step it produced further 'just' to I/O errors. I succesfully mounted this CD and copied files on it (full) and deleted them without errors. After I mounted a DirectCD-formatted disk, I faced with a bounch of WRITE errors, and an irq timeout followed by an ATAPI reset. The best of all: I could use the Linux-formatted CDRW disk by DirectCD and InCD! (Some DirectCD formatted disk can't be used on specific InCD machines...) I suspect, that my writer may not be reliable at 4x speed. The kernel recognise it as a 4x rewriter. The disks used are brand new (Linux-formatted) or known to be good (Windows-formatted), and capable of 4x speed. I also met a problem with my new LinuxUDF CDRW. I copied some file using DIrectCD by WinCommander from a CD. The CD contained old-DOS fashioned files and directories (8.3, all uppercase). I've also copied some other files on the disk at the same time. I can see the copied CD contents only under Windows, while the other files can also be seen on Linux. My questions: 1. Why is it, that the windows formats the disk slightly bigger? Size does not matter for me, but it is a clear difference I've found between the disks. It doesn't bother me, if it is not related to the above presented problems. 2. What can be the reason, that quick-setuped disk couldn't be mounted, but mkudffs-formatted could? 3. What sould I do with these I/O errors (and also with WRITE and read errors possibly related to them)? May I ignore the I/O errors during mkudffs, because the resulted disk seems to be fine? 4. Should I specify lower writing speed somehow, or I should cure my paranoidity? 5. Maybe there are some timing issues with this drive related to the irq handling and so. What should I do? 6. What can be a reason my latter presented problem? Feature request: 1. The possibility to set writing speed explicitly. Thank you rof your attention, Attila Toth --- Attached files: 1. screen-sr0 : screen output using ide-scsi with /* comments */ 2. kernel-sr0 : kernel log digest using ide-scsi with /* comments */ 3. screen-hde : screen output using ide-cd with /* comments */ 4. kernel-hde : kernel log digest using ide-cd with /* comments */
On Thu, 19 Sep 2002, Toth Attila wrote:
I'm using an Acer 4x4x32x IDE-ATAPI CDRW connected to the PIIX4 ide controller of my Abit BP6 motherboard (it also has a hpt366). Let me mention, that this writer shows some strange symtoms: firmware updater programs haven't been able to find the writer on the ide bus for some time
I'm not sure if this is related, but I once owned an USB Acer CDRW that didn't work very well for packet writing until I managed to find a working firmware update. My symptoms were scsi resets, but I don't remember the details any more. Anyway, if you could figure out a way to make the firmware update work, maybe that would solve your problems. -- Peter Osterlund - petero2@telia.com http://w1.894.telia.com/~u89404340
Is there a way to do it? No problem, if the kernel must be recompiled. Thanks, Dw.
On Sat, 5 Oct 2002, Toth Attila wrote:
Is there a way to do it? No problem, if the kernel must be recompiled.
Only if you change the source code too. In drivers/block/pktcdvd.c, the pkt_set_speed function defines the ratio between read speed and write speed. (Currently 3/2.) The pkt_adjust_speed function is called in two places to request a write speed. When opening the device for writing, the code requests 16x speed (in pkt_open_write), and when opening for reading, the code requests maximum speed (in pkt_open_dev). However, when I experimented with this some time ago, I couldn't make it work the way I wanted (dynamic read speed.) It looked like my CDRW ignored the speed setting commands and used whatever speed it thought was best. -- Peter Osterlund - petero2@telia.com http://w1.894.telia.com/~u89404340
participants (2)
-
Peter Osterlund
-
Toth Attila