On Thu, 2007-11-29 at 16:41 -0500, Stephen Smalley wrote: > On errors during libsepol processing (missing requirements, > neverallow failure, etc), we are getting a bogus "Could not allocate memory" > message appended to the libsepol error message by libsemanage when > it reports strerror(errno), which was added to provide useful messages > on e.g. filesystem full or read-only filesystem errors. It appears that > errno is getting set by glibc upon some memory allocations performed by > libsepol even when it ultimately succeeds. So for now, clear such errnos > from libsepol - on real out of memory conditions, libsepol already tells > us that in its own error message. > > Signed-off-by: Stephen Smalley <sds@xxxxxxxxxxxxx> I'm going to NAK my own patch ;) I'll investigate a better approach to handling this, as losing the errno doesn't seem good when there is truly an ENOMEM condition. I'm still not 100% clear as to what precisely is setting it to ENOMEM, as we aren't getting any actual memory failures. > > --- > > libsemanage/src/debug.c | 4 ++++ > 1 file changed, 4 insertions(+) > > Index: trunk/libsemanage/src/debug.c > =================================================================== > --- trunk/libsemanage/src/debug.c (revision 2692) > +++ trunk/libsemanage/src/debug.c (working copy) > @@ -103,6 +103,10 @@ > if (!sh->msg_callback) > return; > > + /* Clear transient memory allocation error, non-fatal. */ > + if (errno == ENOMEM) > + errno = 0; > + > va_start(ap, fmt); > vsnprintf(buffer, sizeof(buffer), fmt, ap); > va_end(ap); > -- Stephen Smalley National Security Agency -- This message was distributed to subscribers of the selinux mailing list. If you no longer wish to subscribe, send mail to majordomo@xxxxxxxxxxxxx with the words "unsubscribe selinux" without quotes as the message.
