|
From: Greg P. <gp...@us...> - 2005-08-24 05:43:10
|
Some potential minor syscall wrapper bugs that I saw while comparing the syscalls to Darwin: * The wrapper for fcntl(F_SETOWN) and fcntl(F_SETSIG) ignore ARG3, but should not. * PRE(sys_select) does in fact sanity-check too little memory for each of the fd_sets (it misses the last n%8 valid bits.) The comment there may indicate that someone already knows about this. * Linux PRE(sys_umount) doesn't print ARG2. -- Greg Parker gp...@us... |
|
From: Nicholas N. <nj...@cs...> - 2005-08-24 13:51:25
|
On Wed, 24 Aug 2005, Greg Parker wrote: > Some potential minor syscall wrapper bugs that I saw while > comparing the syscalls to Darwin: > > * The wrapper for fcntl(F_SETOWN) and fcntl(F_SETSIG) ignore ARG3, > but should not. > > * Linux PRE(sys_umount) doesn't print ARG2. I've fixed these two, thanks. > * PRE(sys_select) does in fact sanity-check too little memory for > each of the fd_sets (it misses the last n%8 valid bits.) The > comment there may indicate that someone already knows about this. Yes. As you say, select can read any number of bits from the final byte, and we don't have that level of granularity. I've previously suggested rounding up in this case, but there was disagreement due to the possibility of false positives. N |