On 07/20/2018 04:57 AM, AKASHI Takahiro wrote:
This patch series is an attempt to address FAT write related issues in an effort of running UEFI SCT (Self-Certification Test) to verify UEFI support on u-boot.
SCT is a test platform as well as an extentisive collection of test cases for UEFI specification. It can run all the tests automatically and save test results to dedicated log files.
AFAIK, what's missing in the current fat file system to safely run SCT without errors (I don't mean test case failures) are:
- write a file located under sub-directories
- write a file with non-zero offset
- delete a file
- create a directory
Patch#1 to patch#6 are some sort of preparatory ones. Patch#7 implements write with sub-directories path. Patch#8 to patch#10 implement write with non-zero offset. Patch#11 to patch#15 are related to creating a directory. Patch#16 provides delete, but it doesn't actually delete a file but instead return 0 with purging a file content, which is good enough to run SCT for now. Finally, patch#17 fixes a minor bug in fs-test.sh.
I applied this patch set on top of v2018.07 along with a couple of yet-to--be-upstreamed UEFI-related patches, and could successfully run unmodified SCT[1] on qemu-arm.
Thanks Takahiro for getting this all implemented.
I have some questions concerning code pages:
The FAT file system contains a short name and a long name. The short name is in the 8-bit code page of the system. The long name is 16bit Unicode.
Which local code page do you assume? I would suggest to use 437 as we do in some other drivers.
With your patch fat_mkdir() and fat_file_write() do not yet support UTF-8 as file names and convert this to the 437 local page for the short name and to UTF-16 for the long name.
Without this conversoin we will not be able to implement the EFI specification.
normalize_longname() simply ignores codes 0x80-0xFF. Please, have a look at
https://www.win.tue.nl/~aeb/linux/fs/fat/fatgen103.pdf Microsoft Extensible Firmware Initiative FAT32 File System Specification FAT: General Overview of On-Disk Format Version 1.03, December 6, 2000
Shouldn't we implement the base-name algorithm as described in the paper?
Best regards
Heinrich