On Thursday, 2 August 2018 09:56:49 CEST Bjoern Walk wrote:
Make the generation of random bits in virRandomBits independent of the endianness of the running architecture.
This also solves problems with the mocked random byte generation on big-endian machines.
Suggested-by: Daniel P. Berrangé <berrange@redhat.com> Signed-off-by: Bjoern Walk <bwalk@linux.ibm.com> --- This goes on top of Michal's fix: https://www.redhat.com/archives/libvir-list/2018-August/msg00080.html
src/util/virrandom.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/src/util/virrandom.c b/src/util/virrandom.c index 7915f653..26ff68f5 100644 --- a/src/util/virrandom.c +++ b/src/util/virrandom.c @@ -34,6 +34,7 @@ # include <gnutls/crypto.h> #endif
+#include "virendian.h" #include "virrandom.h" #include "virthread.h" #include "count-one-bits.h" @@ -61,13 +62,16 @@ VIR_LOG_INIT("util.random"); uint64_t virRandomBits(int nbits) { uint64_t ret = 0; + uint8_t val[8];
- if (virRandomBytes((unsigned char *) &ret, sizeof(ret)) < 0) { + if (virRandomBytes((unsigned char *) &val, sizeof(val)) < 0) { /* You're already hosed, so this particular non-random value * isn't any worse. */ return 0; }
+ ret = virReadBufInt64LE(val); +
I do not think this patch is correct: we are dealing with random bytes, so there is no "endianness" for them. -- Pino Toscano