On Sat, 29 Jul 2006, Linus Torvalds wrote: > > Are you _really_ suggesting that people who happen to log in to the > console should automatically also be allowed to rewrite the CD-ROM drive > firmware, just because they are supposed to be able to write to the CD-ROM > media in that drive? Or do other random things that the kernel really > doesn't know what they do? Btw, in the name of full disclosure, I should probably admit that when I did some of the block/scsi_ioctl.c reorganizations and tried to push people to realize that it was much better to do "cdrecord dev=/dev/hda" than by going through sg.c and the ide-scsi emulation thing, the code not only only first missed the CAP_SYS_RAWIO check entirely, I also then made the decision that only root could do any SCSI command ioctl's, and that I thought it was reasonable. I was disabused of that notion fairly quickly. People _really_ didn't like the "if (!capable(CAP_SYS_RAWIO)) return -EPERM;" approach, and "verify_command()" was written pretty quickly. So I have been disabused of that notion myself, the same way I now hope to pass on that wisdom to the next generation of abusees ;) Anyway, the kernel _has_ gone both ways in the past - not having any access checks at all, and only allowing root. We tried it, and neither really worked. And no, to my knowledge, nobody _actually_ ever fried their CD-ROM by a bad user overwriting the firmware, but quite frankly, I don't want to risk it. (As a historical oddity - there _was_ this bug in some drive that would corrupt the firmware by mistake, though - I think we _did_ actually fry some drives by sending it a "cache flush" command that it wasn't expecting, and that it turned into firmware reload thing or something. I think Alan Cox knows all the gruesome details, but that we could definitely blame on bad hardware.. I think that actually happened to be a CD-ROM drive, but it wasn't even a user-generated command at all, it was the kernel doing the cache-flush all on its sorry own.. Somebody who remembers the exact details can fill us in just for completeness) Linus - : send the line "unsubscribe linux-scsi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
