PANIC: vm_page_fault: unhandled page fault in kernel space at 0x24, ip 0x80039a50

[jonas.kirilla]

See attached file.

Reproducible. Happens with hrev35136.

[jonas.kirilla]

FWIW, the filename is "[haiku-commits] hrev35066 - in haiku_trunk: headers_private_kernel_arch headers_private_kernel_arch_ppc headers_private_kernel_arch_x86 headers_private_kernel_vm src_system_kernel_arch_generic ... 20100114042613 ingo_weinhold"

[anevilyak]

The problem appears to be as follows: Inode::ReadAt() calls file_cache_read() with a NULL file cache pointer (which it is not prepared to cope with as it triggers the invalid fault while reading ref->disabled_count), because Inode only creates a cache for files and attributes, which a symlink doesn't match. However, what I don't understand is how this doesn't panic on every symlink.

[anevilyak]

Upon closer inspection, I see why this particular one triggers it: It's long enough for INODE_LONG_SYMLINK to be set in its flags, which triggers the inode read. Otherwise the read isn't needed.

[anevilyak]

Proposed patch

[anevilyak]

Attached a (confirmed working) patch, please review when you get a chance Axel :)

[anevilyak]

Fixed in hrev35147.