Re: [libvirt] PATCH: Support storage copy on write volumes

Jim Meyering píše v Čt 22. 01. 2009 v 19:15 +0100: > > +static int > > +cowGetBackingStore(virConnectPtr conn, > > + char **res, > > + const unsigned char *buf, > > + size_t buf_size) > > +{ > > + size_t len; > > + > > + *res = NULL; > > + if (buf_size < 4+4+1024) > > + return BACKING_STORE_INVALID; > > + if (buf[4+4] == '\0') /* cow_header_v2.backing_file[0] */ > > + return BACKING_STORE_OK; > > + > > + len = 1024; > > + if (VIR_ALLOC_N(*res, len + 1) < 0) { > > + virStorageReportError(conn, VIR_ERR_NO_MEMORY, _("backing store path")); > > + return BACKING_STORE_ERROR; > > + } > > + memcpy(*res, buf + 4+4, len); /* cow_header_v2.backing_file */ > > + (*res)[len] = '\0'; > > + if (VIR_REALLOC_N(*res, strlen(*res) + 1) < 0) { > > Is this just-copied 1024-byte block of data guaranteed > not to contain any NUL bytes? Or maybe you just want that > NUL-terminated string?

The 1024 bytes should contain a NUL-terminated string. This code ensures there is a trailing NUL, then resizes *res to only contain the first NUL-terminated string. We could avoid the realloc() by using strnlen() from gnulib.