2.4.X includes both read/write support for UDF based on the 0.8.9.X release.
The latest version is 0.9.3. The 0.9 series made some major write changes, which is why I havn't sent a diff off to Linus.
If you want to use UDF, I'd suggest grabbing the cvs source off of sourceforge and just compiling it as a module. (linux-udf project)
For DVD-RAM, as long as you compile UDF with write support, you should be able to mount it and write to it (after making the filesystem on it - or using a disc with a pre-existing filesystem on it).
For CD-RW, you need packet writing support. I recently gave a shot at porting pktcdvd from Jens Axboe to 2.4.2 and I got it working, but its not especially stable. =) For information on pktcdvd check out the packet-cd project on sourceforge.
Packet writing is only needed for CD-RW drives. With DVD-RAM, you can just write away. There were some patches for 2.2 which made the DVD-RAM drive show up as 2 devices. 1 was a cdrom, and 1 was a harddisk. In 2.4, support was added for writing to the cdrom driver, so it should only show up as one device, but it should be writeable (at least thats how I understand things should work).
The major issues with CDRWs (above and beyond UDF write support) are: 1. CDRW drives don't do bad block remapping. So all bad block remapping has to be handled by the filesystem (UDF supports this). Of course, the drive also doesn't tell you when the write failed. So you have to read back the block to see if the write succeeded.... BUT
2. You can't just write blocks to a CDRW (well, ok.. you can.. BUT.. writing to CDRWs consists of writing packets.. now, your packet length could be 1 (in which case, your writing blocks to the media) but for each packet, you incure a 7 block overhead (4 blocks lead-in, 3 blocks lead-out or something along those lines). So, take 650 MB, divide by 8, and you get the size of the disk written with 1 block packets. Not very useful:)
3. Soo.. The packet size used is 32. Which means, when you write to the media, you have to write 32 blocks, or 64k.. Of course, as far as I know, the kernel has no support for doing this (the scsi generic module doesn't even allocate a large enough buffer to support it in userspace).
4. But remember, you gotta check if all the blocks in the packet were successfully written.. So after writing the 64k chunk, you have to read it back.. Fun!
5. And last but not least, unless you have a really new CD drive (and even then) you might not be able to read a fixed packet written disk in anything but your CDRW. If your looking for a new CD drive, you want a drive that handles both fixed and variable packet writen media.
I think I got all the gotchas for CDRWs. At least, as far as my knowledge goes. If I said something totally wrong, let me know:)
Slashdot Mirror
Ok, here's whats up =)
2.4.X includes both read/write support for UDF based on the 0.8.9.X release.
The latest version is 0.9.3. The 0.9 series made some major write changes, which is why I havn't sent a diff off to Linus.
If you want to use UDF, I'd suggest grabbing the cvs source off of sourceforge and just compiling it as a module. (linux-udf project)
For DVD-RAM, as long as you compile UDF with write support, you should be able to mount it and write to it (after making the filesystem on it - or using a disc with a pre-existing filesystem on it).
For CD-RW, you need packet writing support. I recently gave a shot at porting pktcdvd from Jens Axboe to 2.4.2 and I got it working, but its not especially stable. =) For information on pktcdvd check out the packet-cd project on sourceforge.
Ben
Packet writing is only needed for CD-RW drives. With DVD-RAM, you can just write away. There were some patches for 2.2 which made the DVD-RAM drive show up as 2 devices. 1 was a cdrom, and 1 was a harddisk. In 2.4, support was added for writing to the cdrom driver, so it should only show up as one device, but it should be writeable (at least thats how I understand things should work).
Ben
The major issues with CDRWs (above and beyond UDF write support) are:
:)
:)
1. CDRW drives don't do bad block remapping. So all bad block remapping has to be handled by the filesystem (UDF supports this). Of course, the drive also doesn't tell you when the write failed. So you have to read back the block to see if the write succeeded.... BUT
2. You can't just write blocks to a CDRW (well, ok.. you can.. BUT.. writing to CDRWs consists of writing packets.. now, your packet length could be 1 (in which case, your writing blocks to the media) but for each packet, you incure a 7 block overhead (4 blocks lead-in, 3 blocks lead-out or something along those lines). So, take 650 MB, divide by 8, and you get the size of the disk written with 1 block packets. Not very useful
3. Soo.. The packet size used is 32. Which means, when you write to the media, you have to write 32 blocks, or 64k.. Of course, as far as I know, the kernel has no support for doing this (the scsi generic module doesn't even allocate a large enough buffer to support it in userspace).
4. But remember, you gotta check if all the blocks in the packet were successfully written.. So after writing the 64k chunk, you have to read it back.. Fun!
5. And last but not least, unless you have a really new CD drive (and even then) you might not be able to read a fixed packet written disk in anything but your CDRW. If your looking for a new CD drive, you want a drive that handles both fixed and variable packet writen media.
I think I got all the gotchas for CDRWs. At least, as far as my knowledge goes. If I said something totally wrong, let me know