On Wed, 2019-07-24 at 02:13 +0300, Vitaly Chikunov wrote:
> Mimi,
>
> On Tue, Jul 23, 2019 at 06:59:16PM -0400, Mimi Zohar wrote:
> > On Wed, 2019-07-24 at 01:18 +0300, Vitaly Chikunov wrote:
> > > On Thu, Jul 18, 2019 at 10:29:54AM -0400, Mimi Zohar wrote:
> > > > Each tima a new unknown key is encountered, emit a message of the format
> > > > "key #: <keyid> unknown". The individual files using unknown keys are
> > > > then only logged in verbose mode.
> > > >
> > > > Signed-off-by: Mimi Zohar <zohar@xxxxxxxxxxxxx>
> > > > ---
> > > > src/libimaevm.c | 21 +++++++++++++++++----
> > > > 1 file changed, 17 insertions(+), 4 deletions(-)
> > > >
> > > > diff --git a/src/libimaevm.c b/src/libimaevm.c
> > > > index 43eb4ef2412c..d2194a6ca0f8 100644
> > > > --- a/src/libimaevm.c
> > > > +++ b/src/libimaevm.c
> > > > @@ -402,13 +402,26 @@ static struct public_key_entry *public_keys = NULL;
> > > >
> > > > static EVP_PKEY *find_keyid(uint32_t keyid)
> > > > {
> > > > - struct public_key_entry *entry;
> > > > + struct public_key_entry *entry, *tail = public_keys;
> > >
> > > If user specified in `-k` filename that does not exist no key is added
> > > into public_keys and it remains NULL.
> >
> > Right, just the keyid is added and an error message is emitted.
> >
> > >
> > > > + int i = 1;
> > > >
> > > > for (entry = public_keys; entry != NULL; entry = entry->next) {
> > > > if (entry->keyid == keyid)
> > > > return entry->key;
> > > > + i++;
> > > > + tail = entry;
> > > > }
> > > > - return NULL;
> > > > +
> > > > + /* add unknown keys to list */
> > > > + entry = calloc(1, sizeof(struct public_key_entry));
> > > > + if (!entry) {
> > > > + perror("calloc");
> > > > + return 0;
> > > > + }
> > > > + entry->keyid = keyid;
> > > > + tail->next = entry;
> > >
> > > In that case here is SIGSEGV when user try to ima_verify.
> >
> > find_keyid() returns NULL, which is checked before being used.
> > There's only one caller of this function.
>
> find_keyid does not return since tail is NULL, thus there is null
> dereference in `tail->next`.
>
> > Do you have a test case to reproduce this bug?
>
> ima-evm-utils/tests ((31b5f50...))$ rm sha1.txt
> ima-evm-utils/tests ((31b5f50...))$ touch sha1.txt
> ima-evm-utils/tests ((31b5f50...))$ evmctl ima_sign -k test-rsa1024.key -a sha1 --xattr-user sha1.txt
> ima-evm-utils/tests ((31b5f50...))$ evmctl ima_verify -k /absent --xattr-user sha1.txt
> Failed to open keyfile: /absent
> Segmentation fault
>
> 31b5f50 is head of your git next branch.
>
> I temporary fixed like this:
>
> - tail->next = entry;
> + if (tail)
> + tail->next = entry;
> + else
> + public_keys = entry;
I see. If you don't object, I'll fold this into the original patch.
Mimi
