crash: Wrong kind of BTree header

[tomkoen]

Just have encounted a DMG file (Apple Service Diagnostic 3S135.dmg, 3.55 GB) that throws this type of exception (in HFSBTree::HFSBTree) when constructing HFSExtentsOverflowBTree
After this, the app crashes in CachedReader::nonCachedRead
throw std::logic_error("Range returned by adviseOptimalBlock() is too large");

[tomkoen]

The DMG to reproduce:
https://www.dropbox.com/s/1si6j71ak4jxny7/Apple%20Service%20Diagnostic%203S135.dmg?dl=0

[LubosD]

I can confirm the issue with adviseOptimalBlock(), but I haven't yet seen the HFSExtentsOverflowBTree problem.

[tomkoen]

I can confirm that
std::logic_error("Range returned by adviseOptimalBlock() is too large"); is fixed, thanks!
but "Wrong kind of BTree header" can be reproduced in this way (sorry not to mention it earlier):

// if (DMGDisk::isDMG(g_fileReader)) // if we skip DMG check
//      g_partitions.reset(new DMGDisk(g_fileReader));
//  else 
if (GPTDisk::isGPTDisk(g_fileReader))
        g_partitions.reset(new GPTDisk(g_fileReader));
    else if (AppleDisk::isAppleDisk(g_fileReader))
        g_partitions.reset(new AppleDisk(g_fileReader));
    else if (HFSVolume::isHFSPlus(g_fileReader)) // this is TRUE!
        volume.reset(new HFSVolume(g_fileReader)); // here is exception
    else
        throw function_not_implemented_error("Unsupported file format");

It might not be an error I understand

[LubosD]

Is is an error, which would normally prevent the image from being mounted on Linux, but it just doesn't happen here. I'll download the file again from your DropBox, maybe it's different?

[tomkoen]

Sorry but it doesn't happen when trying to mount the file as DMG.
For some reason HFSVolume::isHFSPlus gives true for this file and then

volume.reset(new HFSVolume(g_fileReader));

gives exception

[LubosD]

Ah, I see. I'll take a look into that. But given that the .DMG file is itself RAW (=no compression used), it is entirely possible it may get misdetected.