I've found something that looks odd in the way v9fs under Linux interacts with a fossil running on Plan 9. If I create a file by redirecting output, the shell seems able to happily create the file. If I try to cp a file, then it fails. The logs show the error "unknown mode." As near as I can tell, here's what's happening:
- cp creates the file with the O_EXCL flag. - v9fs_uflags2mode() sets the P9_OEXCL flag. - v9fs defines P9_OEXCL to be 4, where /sys/include/libc.h defines OEXCL to be 0x1000 - fossil masks the mode it gets with 7, keeping the P9_OEXCL flag. - When fossil checks for the allowed modes, it does so with strict equalities that don't include bit 2 (the P9_OEXCL flag). So none of them match, and we get an unknown mode. At first glance, it would appear that there are two bugs. v9fs shouldn't define P9_OEXCL to be 4, but to be 0x1000. Second the OMODE mask in /sys/src/cmd/fossil/9p.c should be 3 and not 7 so that we are only looking at the read and write flags when doing the comparisons. Am I missing something here? Is there a reason these are what they are? If they've been identified and fixed already, I apologize for the noise. BLS
