Em sex., 16 de jan. de 2026 às 13:55, Ján Tomko <[email protected]> escreveu: > > On a Monday in 2026, Julio Faracco wrote: > >The virSocketAddrFormatWithPrefix() function has a bug where the > >'network' variable is left uninitialized when masked=false. This > >occurs because the function only assigns to 'network' inside the > >masked=true conditional branch. > > > >When masked=false, the caller wants to format the original address > > There is no such caller, ever since its introduction in: > commit 426afc0082f1d28449380a5c9260d64ed7183e38 > util: rename/move iptablesFormatNetwork to virSocketAddrFormatWithPrefix > we always passed masked=true. > > I think dropping the "masked" argument is easier here.
I was thinking of renaming the function name to something like: virSocketAddrFormatWithMask and drop the argument, but I would like to see opinions first. Seems the right way (drop the argument) based on your comments and context. > > Also, calling it "unitialized" evokes some kind of omission that made > the function work by accident. Here, the "addr" is never used > so the function would not even work. > > Jano > > >with a prefix notation (e.g., "1.2.3.4/24") without applying the > >network mask. However, the code was only initializing 'network' > >when masking was requested, causing the subsequent > >virSocketAddrFormat(&network) call to operate on uninitialized data. > > > >Fix this by adding an else branch that copies the original address > >to 'network' when masking is not requested. This ensures 'network' > >is properly initialized in both code paths. ACK. Submitting a V2 with masked code only makes more sense. > > > >Signed-off-by: Julio Faracco <[email protected]> > >--- > > src/util/virsocketaddr.c | 12 ++++++++---- > > 1 file changed, 8 insertions(+), 4 deletions(-)
