Tails

#1 Updated by intrigeri 2014-06-25 11:03:30

> isohybrid tails-i386-1.1~beta1.iso --entry 4 --type 0x1c returns:

> isohybrid: Warning: more than 1024 cylinders: 1045
> isohybrid: Not all BIOSes will be able to boot this device

What’s the version of isohybrid being used?

Are these messages mere warnings, or fatal errors? In other words: 1. once the ISO is cat’ed on a device, does it actually boot? 2. is the ISO file modified in the process?

#4 Updated by intrigeri 2014-06-25 12:15:31

Maybe we could simply redirect stdout/stderr/whatever to /dev/null.

#8 Updated by BitingBird 2014-07-30 23:00:06

Well, I have some users who have this error and the USB doesn’t boot.

#11 Updated by intrigeri 2014-07-31 14:03:14

The tests run as part of Feature #7692 showed that these warnings are caused by the increase of the size of our ISO. Now, it might be that some other combination of options passed to isohybrid break the boot on less systems. Hence:

1. sent a call for testing on -testers@ and -dev@;
2. asked for advice on the syslinux mailing-list.

#12 Updated by intrigeri 2014-07-31 17:32:39

What I gathered already:

• --entry 4 --type 0x1c was meant to support BIOS’es that only boot from USB in USB-Zip mode; these BIOS’es were said to be ancient… in 2009; presumably, systems with such a BIOS can’t boot Tails anyway;
• -h 255 -s 63 is said to be the second best combination of options, and it’s needed for > 1 GiB ISO images. Indeed, with these options, isohybrid doesn’t print any warning when run on the Tails 1.1 ISO. I’m said that this allows “to get the maximum cylinder size of a bit less than 8 MiB. This will suffice for ISOs up to nearly 8 GiB”.

#14 Updated by intrigeri 2014-07-31 19:01:01

I think we should:

1. Short-term workaround: modify the doc to suggest the aforementioned options (replacing --entry 4 --type 0x1c with -h 255 -s 63) in case the current ones don’t work.
2. Then:
   #* if the feedback we get from the call for testing and/or my reasoning above makes us confident enough to change the isohybrid options, then do it;
   #* else, get Feature #7706 fixed in 1.1.1 to workaround the problem.

#15 Updated by intrigeri 2014-08-01 12:57:40

> Short-term workaround: modify the doc to suggest the aforementioned options (replacing --entry 4 --type 0x1c with -h 255 -s 63) in case the current ones don’t work.

I had a look, and it doesn’t seem easy to adapt the manual installation doc this way: once one has followed the documented steps, their ISO is modified and can’t be used to test another combination of isohybrid options. As a short-term fix (unless we’re bold enough to change the isohybrid options right ahead in the doc, without waiting for testing feedback, which I’m almost tempted to do), maybe Tails does not start could advice to download a fresh ISO, and follow the manual installation process again, except the changed isohybrid options.

sajolida, what do you think?

#16 Updated by sajolida 2014-08-01 15:07:17

> Well, I have some users who have this error and the USB doesn’t boot.

I haven’t identified things like this on tails-bugs@ since the release.

But quite a few people had problems booting Tails 1.1 due to a bad
download or a regression in the support of their graphic card. That
could as well be the case here.

So, how far did you investigate the problem on those USB sticks not
booting? What boot problem where they experiencing?

#17 Updated by intrigeri 2014-08-01 15:40:37

> I haven’t identified things like this on tails-bugs@ since the release.

Yes, that was also my impression, and it made me unsure that we had a problem in the beginning. Just for the record: tails-bugs@ sees less traffic than IRC.

> So, how far did you investigate the problem on those USB sticks not booting? What boot problem where they experiencing?

People doing intensive support on IRC, such as bitingbird, will know better than me.

All that I was told is something along the lines of “for quite a few users, only DVD + clone’n’install works”, which tends to confirm that there’s a problem with isohybrid, and is no big surprise considering 1. the warning it displays; and 2. the replies I got on the syslinux mailing-list from experts in this area (“-h 255 -s 63 is said to be the second best one. Needed for > 1 GiB”), that personally I won’t argue against.

Given all this, I personally need neither more info from bug reporters, nor to see the problem myself, to be convinced that we actually do have a problem.

#18 Updated by sajolida 2014-08-02 10:22:33

> I had a look, and it doesn’t seem easy to adapt the manual installation doc this way: once one has followed the documented steps, their ISO is modified and can’t be used to test another combination of isohybrid options. As a short-term fix (unless we’re bold enough to change the isohybrid options right ahead in the doc, without waiting for testing feedback, which I’m almost tempted to do), maybe Tails does not start could advice to download a fresh ISO, and follow the manual installation process again, except the changed isohybrid options.
>
> sajolida, what do you think?

And also after doing isohybrid people can’t verify the ISO image again
anymore. So what about advising them to do a backup of the ISO image, and
explain that it will be modified by this command?

#19 Updated by intrigeri 2014-08-02 11:23:46

> And also after doing isohybrid people can’t verify the ISO image again
> anymore. So what about advising them to do a back of the ISO image, and
> explain that it will be modified by this command?

Yes. That’s related to Bug #7370.

#20 Updated by intrigeri 2014-08-03 17:16:08

FTR, I’ve just nominated Feature #7706 for 1.1.1, and hope Alan will merge it tomorrow, in the hope that it fixes the root cause of this problem as a stop-gap for now: I’ve still received no test reports regarding the new isohybrid options, so I’m less and less sure we’ll find a good solution to this problem via documentation changes in time for 1.1.1. (Still, documenting workarounds now would still make sense IMO.)

#28 Updated by intrigeri 2014-08-15 10:36:34

Mitigated with Feature #7706 being done for 1.1.1, but that’s only a short-term fix: I wouldn’t be surprised if we hit the 1GB limit by the end of the year, for some reason.

Pinged on the call for testing wrt. new isohybrid options, no answer so far… if we get no more testing results, then I think we should go ahead and change the doc anyway at some point. I’ll come back to it in a week or two.

#30 Updated by intrigeri 2014-08-20 12:04:12

> In case we finally decide to replace the old options with the new ones, I updated the branch doc/7462-new-isohybrid-options to do that.

Great, thanks! I’ve looked at it, and it looks good.

> I’m now reassign this to you either to bring more info, either to review it.

OK. I’ll wait a few more days, in the hope that my last ping on the call for testing finally produces results, and if I see no regression reported, then I’ll merge into master.

#31 Updated by intrigeri 2014-08-25 00:49:07

Test report on macbook pro 9,2: no regression with syslinux 3:6.03~pre18+dfsg-1~bpo70+1. The old options worked fine already.