| Message ID | 20200802171247.16558-1-trix@redhat.com (mailing list archive) |
|---|---|
| State | Rejected |
| Delegated to: | Herbert Xu |
| Headers | show |
| Series | crypto: drbg: check blocklen is non zero | expand |
Am Sonntag, 2. August 2020, 19:12:47 CEST schrieb trix@redhat.com: Hi Tom, > From: Tom Rix <trix@redhat.com> > > Clang static analysis reports this error > > crypto/drbg.c:441:40: warning: Division by zero > padlen = (inputlen + sizeof(L_N) + 1) % (drbg_blocklen(drbg)); > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~ > > When drbg_bocklen fails it returns 0. > > if (drbg && drbg->core) > return drbg->core->blocklen_bytes; > return 0; > > In many places in drbg_ctr_df drbg_bocklen is assumed to be non zero. > So turn the assumption into a check. > > Fixes: 541af946fe13 ("crypto: drbg - SP800-90A Deterministic Random Bit > Generator") > > Signed-off-by: Tom Rix <trix@redhat.com> Thank you. Reviewed-by: Stephan Mueller <smueller@chronox.de> Ciao Stephan
On Sun, Aug 02, 2020 at 10:12:47AM -0700, trix@redhat.com wrote: > From: Tom Rix <trix@redhat.com> > > Clang static analysis reports this error > > crypto/drbg.c:441:40: warning: Division by zero > padlen = (inputlen + sizeof(L_N) + 1) % (drbg_blocklen(drbg)); > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~ > > When drbg_bocklen fails it returns 0. > > if (drbg && drbg->core) > return drbg->core->blocklen_bytes; > return 0; Yes but it can only fail if the drbg is not instantiated. If you're hitting the generate path with an uninstantiated drbg you've got bigger problems than a divide by zero. So how is this even possible? Cheers,
There are many divide by 0 reports. This one got attention because it is in crypto, where i believe problems, even false positives, should be fixed. Tom On 8/20/20 12:15 AM, Herbert Xu wrote: > On Sun, Aug 02, 2020 at 10:12:47AM -0700, trix@redhat.com wrote: >> From: Tom Rix <trix@redhat.com> >> >> Clang static analysis reports this error >> >> crypto/drbg.c:441:40: warning: Division by zero >> padlen = (inputlen + sizeof(L_N) + 1) % (drbg_blocklen(drbg)); >> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~ >> >> When drbg_bocklen fails it returns 0. >> >> if (drbg && drbg->core) >> return drbg->core->blocklen_bytes; >> return 0; > Yes but it can only fail if the drbg is not instantiated. If > you're hitting the generate path with an uninstantiated drbg you've > got bigger problems than a divide by zero. > > So how is this even possible? > > Cheers,
On Thu, Aug 20, 2020 at 06:27:36AM -0700, Tom Rix wrote:
> There are many divide by 0 reports. This one got attention because it is in crypto, where i believe problems, even false positives, should be fixed.
Please don't top post.
AS your bug report is a false positive I'm rejecting your patch.
Cheers,
diff --git a/crypto/drbg.c b/crypto/drbg.c index e99fe34cfa00..bd9a137e5473 100644 --- a/crypto/drbg.c +++ b/crypto/drbg.c @@ -420,6 +420,9 @@ static int drbg_ctr_df(struct drbg_state *drbg, size_t inputlen = 0; struct drbg_string *seed = NULL; + if (!drbg_blocklen(drbg)) + return -EINVAL; + memset(pad, 0, drbg_blocklen(drbg)); memset(iv, 0, drbg_blocklen(drbg));