=== asac_ is now known as asac | ||
=== pgraner is now known as pgraner_afk | ||
lamont | May 21 09:18:29 mix kernel: [159411.675112] rtc: lost 17 interrupts | 17:05 |
---|---|---|
lamont | make it stop saying that, mk? | 17:06 |
psusi | does anyone know why the kernel was changed to default to defeating host protected areas on disks and why this should not be considered a bug and reverted? | 19:38 |
mjg59 | psusi: Because the alternative was for people's systems to stop working when they switched from IDE to libata | 19:49 |
psusi | mjg59: eh? | 19:49 |
mjg59 | Turns out that most people consider that a bug | 19:49 |
psusi | mjg59: I'm seeing people's systems breaking because of this change | 19:49 |
mjg59 | psusi: The IDE system deactivated the HPA since, well, forever | 19:49 |
psusi | oh really? | 19:49 |
mjg59 | Yes | 19:50 |
mjg59 | So people created partitions in the HPA | 19:50 |
mjg59 | When libata didn't deactivate the HPA, stuff broke | 19:50 |
psusi | hrm.... why is that not broken? HPA means you arent supposed to muck with it | 19:50 |
mjg59 | Because their filesystems now extended into unreachable areas of the disk | 19:50 |
psusi | well, yea... but it was broken before and then fixed | 19:50 |
mjg59 | No | 19:50 |
psusi | so we revert to the broken heavior? doesn't seem right | 19:50 |
psusi | behavior even | 19:51 |
mjg59 | There is no way in the universe you can claim that breaking people's filesystems is a fix | 19:51 |
psusi | no, the fix was NOT breaking HPA | 19:51 |
mjg59 | There is no way in the universe you can claim that breaking people's filesystems is a fix | 19:51 |
psusi | but if they already had a broken system then fixing it broke their system | 19:51 |
mjg59 | That's not an acceptable option | 19:52 |
mjg59 | Which is why upstream behaiour was changed | 19:52 |
mjg59 | Also, this happened over a year ago | 19:52 |
psusi | how is it acceptable to allow them to break their system by overwriting the HPA? | 19:52 |
psusi | there may be parts of the bios written there that if we overwrite, the system won't boot | 19:53 |
mjg59 | Yes, it's acceptable to let people do that | 19:53 |
mjg59 | No, there will not be parts of the BIOS written there | 19:53 |
psusi | there could be... you don't know why the bios protected it | 19:53 |
mjg59 | Yes, and it could also contain child pornography | 19:54 |
psusi | it seems that some vendors put bios extension programs there | 19:54 |
mjg59 | In which case being able to delete it would be a good thing | 19:54 |
psusi | the point is that via HPA, the bios is telling us DON'T TOUCH | 19:54 |
mjg59 | No, vendors do not put bios extension programs there | 19:54 |
psusi | if you as the user want to override that, that's one thing... but doing it by default? | 19:55 |
psusi | I have heard of some that do | 19:55 |
psusi | there is at least one vendor I have heard of that uses it to store some sort of instant on cd playing program | 19:55 |
psusi | so you can play audio cds without booting up fully | 19:55 |
mjg59 | Yes, it's a filesystem containing Linux | 19:55 |
mjg59 | It's not a BIOS extension | 19:56 |
psusi | right... but it's marked as protected... so then we come and install Ubuntu and trash it | 19:56 |
mjg59 | So don't trash it | 19:56 |
mjg59 | The partitioner makes it easy to avoid this | 19:56 |
psusi | kind of hard when it just looks like the tail end of free space on the disk | 19:56 |
mjg59 | No, it'll still look like a partition | 19:56 |
psusi | it isn't listed in the partition table since it exists in the hidden part of the disk | 19:56 |
mjg59 | jpgpls | 19:57 |
psusi | though at least in this one person's case, the protected area doesn't appear to contain anything.... but for some reason the bios wanted it protected and it's causing a problem when the disk size does not match what the bios recorded, which was minus the hpa | 19:58 |
mjg59 | "disk size"? | 19:59 |
mjg59 | If there's an HPA, the BIOS will still see the HPA | 20:00 |
mjg59 | It's not altered until Linux starts up | 20:00 |
psusi | it appears to record the size without the hpa when it writes the fakeraid metadata | 20:00 |
mjg59 | Hm. Though, potentially, it would cause problems if the kernel or grub files end up in the HPA | 20:00 |
psusi | then linux starts, disables the hpa, and suddenly the disk size is wrong | 20:00 |
mjg59 | psusi: At which point the BIOS isn't called again, so there's no problem | 20:00 |
psusi | yea, there is... the metadata is now in the wrong location | 20:01 |
mjg59 | What metadata? | 20:01 |
mjg59 | What do you mean by "wrong location"? | 20:01 |
psusi | it's supposed to be the last 2 sectors of the disk... but in this case, it's actually about a meg short of there | 20:01 |
psusi | fakeraid metadata | 20:01 |
mjg59 | What? | 20:01 |
mjg59 | Ah | 20:01 |
psusi | the bios writes it to what IT sees as the last 2 sectors | 20:01 |
psusi | but that's just before the HPA starts | 20:01 |
mjg59 | That's more interesting, but easy to fix | 20:01 |
psusi | we lift the HPA and now it's not in the last two sectors | 20:02 |
mjg59 | Ridiculously retarded BIOS | 20:02 |
psusi | probably ;) | 20:02 |
psusi | "but it works fine with windows" | 20:02 |
mjg59 | But, as I said, easy to fix | 20:02 |
mjg59 | The fakeraid code needs to check both the reported size and the old size | 20:02 |
psusi | ack | 20:03 |
psusi | hrm.... how would it find the old size? | 20:03 |
mjg59 | The kernel has the original size | 20:04 |
mjg59 | It reads it before messing with the HPA | 20:04 |
psusi | any idea how one would look that up? | 20:05 |
mjg59 | In-kernel? Not off-hand. | 20:05 |
mjg59 | Just grep for hpa in drivers/ata/ | 20:05 |
psusi | hrm... guess it's time to do a git clone | 20:05 |
mjg59 | ANyway, getting into the station now | 20:05 |
* mjg59 leaves | 20:05 | |
psusi | mjg59: it appears that the kernel does NOT retain the HPA disk size after lifting the limit | 21:59 |
psusi | libata defines the parameter ata_ignore_hpa... can you not just pass ata_ignore_hpa=0 on the kernel command line if it is built as a module? | 22:06 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!