If we hit an unknown header type (likely to be a corrupt record), and

the user choose to not abort, skip to the next header instead of trying
to use it.
This allowed me to recover files from a corrupt dump, instead of
getting a segfault.
This commit is contained in:
bouyer 2020-09-03 19:31:34 +00:00
parent ed3d9167a3
commit 33be3b5b89

View File

@ -1,4 +1,4 @@
/* $NetBSD: tape.c,v 1.68 2015/03/02 03:17:24 enami Exp $ */
/* $NetBSD: tape.c,v 1.69 2020/09/03 19:31:34 bouyer Exp $ */
/*
* Copyright (c) 1983, 1993
@ -39,7 +39,7 @@
#if 0
static char sccsid[] = "@(#)tape.c 8.9 (Berkeley) 5/1/95";
#else
__RCSID("$NetBSD: tape.c,v 1.68 2015/03/02 03:17:24 enami Exp $");
__RCSID("$NetBSD: tape.c,v 1.69 2020/09/03 19:31:34 bouyer Exp $");
#endif
#endif /* not lint */
@ -1384,6 +1384,7 @@ findinode(struct s_spcl *header)
top:
do {
if (header->c_magic != FS_UFS2_MAGIC) {
skip:
skipcnt++;
while (gethead(header) == FAIL ||
header->c_date != dumpdate)
@ -1438,7 +1439,8 @@ findinode(struct s_spcl *header)
default:
panic("unknown tape header type %d\n", spcl.c_type);
break;
fprintf(stderr, "skiping to next header\n");
goto skip;
}
} while (header->c_type == TS_ADDR);