Use "read" to fetch misaligned 64bit values from bundle header #63431
Conversation
|
Tagging subscribers to this area: @agocke, @vitek-karas, @VSadov Issue DetailsFixes: #62273 Misalligned reads may cause crashes on ARM32, depending on system config.
|
VSadov
changed the title
| if (bundle_major_version >= 6) | ||
| { | ||
| reader.read(&fixed_data.compressedSize, sizeof(int64_t)); | ||
| } | ||
|
|
||
| fixed_data.type = *(file_type_t*)reader.read_direct(sizeof(file_type_t)); |
There was a problem hiding this comment.
What is the size of file_type_t here and how is it guaranteed to be aligned? or do reads of its size not need aligning?
There was a problem hiding this comment.
It is an enum with base uint8_t, defined here:
src/native/corehost/bundle/reader.h
Outdated
| // Return a pointer to the requested bytes within the memory-mapped file. | ||
| // Skip over len bytes. | ||
| const char* read_direct(int64_t len) | ||
| uint8_t read_byte() |
There was a problem hiding this comment.
Nit: since this is only used from one place, is it worth adding? Why not just use the read even for the single byte file type?
There was a problem hiding this comment.
Just noticed the same thing - there is read(), which is the same as read_byte()
|
I've removed |
jeffschwMSFT
removed
the
Servicing-consider
|
Just FYI. I have got Raspbian installed on my PI and I was able to reproduce the failure. (after setting cpu/alignment handler to 4) Now I am at the stage where I want to make sure that the fix actually fixes the problem - i.e. whether after the fix there is no Bus error and apps are able to run. |
|
I have validated that an app built with singlefilehost that contains the fix runs on Raspbian (bullseye arm32) regardless of |
|
/backport to release/6.0 |
|
Started backporting to release/6.0: https://github.com/dotnet/runtime/actions/runs/1669737088 |
ghost
locked as resolved and limited conversation to collaborators
Fixes: #62273
Misalligned reads may cause crashes on ARM32, depending on system config.