|
From: Richard W. <ri...@no...> - 2015-11-28 21:40:37
|
Am 28.11.2015 um 22:32 schrieb Mickaël Salaün: > Replace the default insecure mode 0777 with 0700 for temporary file. > > Prohibit other users to change the executable mapped code. Hmm, isn't the tmp file already unlinked at this stage? Thanks, //richard |
|
From: Richard W. <ri...@no...> - 2015-11-28 22:07:19
|
Am 28.11.2015 um 22:32 schrieb Mickaël Salaün: > Open the memory mapped file with the O_TMPFILE flag when available. > > Signed-off-by: Mickaël Salaün <mi...@di...> > --- > arch/um/os-Linux/mem.c | 12 ++++++++++++ > 1 file changed, 12 insertions(+) > > diff --git a/arch/um/os-Linux/mem.c b/arch/um/os-Linux/mem.c > index 798aeb4..fe52e2d 100644 > --- a/arch/um/os-Linux/mem.c > +++ b/arch/um/os-Linux/mem.c > @@ -106,6 +106,18 @@ static int __init make_tempfile(const char *template) > } > } > > +#ifdef O_TMPFILE > + fd = open(tempdir, O_CLOEXEC | O_RDWR | O_EXCL | O_TMPFILE, 0700); > + /* > + * If the running system does not support O_TMPFILE flag then retry > + * without it. > + */ > + if (fd != -1 || (errno != EINVAL && errno != EISDIR && Why are you handling EISDIR? > + errno != EOPNOTSUPP)) > + return fd; > + errno = 0; Why are you resetting errno? Thanks, //richard |
|
From: Richard W. <ri...@no...> - 2015-11-28 22:55:33
|
Am 28.11.2015 um 23:52 schrieb Mickaël Salaün: > > On 28/11/2015 22:40, Richard Weinberger wrote: >> Am 28.11.2015 um 22:32 schrieb Mickaël Salaün: >>> Replace the default insecure mode 0777 with 0700 for temporary file. >>> >>> Prohibit other users to change the executable mapped code. >> >> Hmm, isn't the tmp file already unlinked at this stage? >> > > Yes, but if someone could open it before the unlink e.g. because of the umask (which does not seems to be the case thanks to mkstemp, but remains unspecified [1]), this user should then be able to have write access to the file descriptor/description. Yes, someone can open it before the unlink. But you change the file mode after that. How does it improve the situation? The attacker has already the file handle. Thanks, //richard |
|
From: Richard W. <ri...@no...> - 2015-11-28 22:59:11
|
Am 28.11.2015 um 23:56 schrieb Mickaël Salaün: > > On 28/11/2015 23:07, Richard Weinberger wrote: >> Am 28.11.2015 um 22:32 schrieb Mickaël Salaün: >>> Open the memory mapped file with the O_TMPFILE flag when available. >>> >>> Signed-off-by: Mickaël Salaün <mi...@di...> >>> --- >>> arch/um/os-Linux/mem.c | 12 ++++++++++++ >>> 1 file changed, 12 insertions(+) >>> >>> diff --git a/arch/um/os-Linux/mem.c b/arch/um/os-Linux/mem.c >>> index 798aeb4..fe52e2d 100644 >>> --- a/arch/um/os-Linux/mem.c >>> +++ b/arch/um/os-Linux/mem.c >>> @@ -106,6 +106,18 @@ static int __init make_tempfile(const char *template) >>> } >>> } >>> >>> +#ifdef O_TMPFILE >>> + fd = open(tempdir, O_CLOEXEC | O_RDWR | O_EXCL | O_TMPFILE, 0700); >>> + /* >>> + * If the running system does not support O_TMPFILE flag then retry >>> + * without it. >>> + */ >>> + if (fd != -1 || (errno != EINVAL && errno != EISDIR && >> >> Why are you handling EISDIR? > > I follow the man page for open [1], I think it was a workaround needed for some kernel versions just after the O_TMPFILE was added but before the support for EOPNOTSUPP. > We may need to add the EACCES too for some version of glibc [2, 3]? Makes sense! :) > 1. http://man7.org/linux/man-pages/man2/openat.2.html#BUGS > 2. Commit 69a91c237ab0ebe4e9fdeaf6d0090c85275594ec and https://sourceware.org/bugzilla/show_bug.cgi?id=17523 > 3. https://bugs.gentoo.org/529044 > >> >>> + errno != EOPNOTSUPP)) >>> + return fd; >>> + errno = 0; >> >> Why are you resetting errno? > > It's to ignore/reset the error code from open, but it may not be needed because of the next call to malloc? But then you'd have to reset errno after every syscall. :-) Thanks, //richard |
|
From: Richard W. <ri...@no...> - 2015-11-28 23:11:48
|
Am 29.11.2015 um 00:00 schrieb Mickaël Salaün: > > > On 28/11/2015 23:55, Richard Weinberger wrote: >> Am 28.11.2015 um 23:52 schrieb Mickaël Salaün: >>> >>> On 28/11/2015 22:40, Richard Weinberger wrote: >>>> Am 28.11.2015 um 22:32 schrieb Mickaël Salaün: >>>>> Replace the default insecure mode 0777 with 0700 for temporary file. >>>>> >>>>> Prohibit other users to change the executable mapped code. >>>> >>>> Hmm, isn't the tmp file already unlinked at this stage? >>>> >>> >>> Yes, but if someone could open it before the unlink e.g. because of the umask (which does not seems to be the case thanks to mkstemp, but remains unspecified [1]), this user should then be able to have write access to the file descriptor/description. >> >> Yes, someone can open it before the unlink. But you change the file mode after that. >> How does it improve the situation? The attacker has already the file handle. > > The attacker could have the file handle only in a read-only mode, which is a bit different than being able to write and execute arbitrary code thanks to a file descriptor mapped RWX :) Fair point. Please describe this in detail in the patch changelog. :-) Thanks, //richard |