From: Ben Gras <beng@shrike-systems.com>
Date: Thu, 2 Apr 2015 11:04:27 +0000 (+0200)
Subject: libc: FTS support for dynamic inode numbering
X-Git-Url: http://zhaoyanbai.com/repos/%22http:/www.isc.org/icons/zpipe.c?a=commitdiff_plain;h=b329f2c73b7e98fad86b1336475663feba53338d;p=minix.git

libc: FTS support for dynamic inode numbering

Edited by Lionel Sambuc and David van Moolenbroek.

Change-Id: I29d6383499d8c0524f86f9dcec701aff35ce8a43
---

diff --git a/lib/libc/gen/fts.c b/lib/libc/gen/fts.c
index d1236fd5f..c5141f1a7 100644
--- a/lib/libc/gen/fts.c
+++ b/lib/libc/gen/fts.c
@@ -1197,6 +1197,10 @@ fts_maxarglen(char * const *argv)
 	return (max + 1);
 }
 
+#ifdef __minix
+#include <minix/dmap.h>
+#endif
+
 /*
  * Change to dir specified by fd or p->fts_accpath without getting
  * tricked by someone changing the world out from underneath us.
@@ -1217,7 +1221,22 @@ fts_safe_changedir(const FTS *sp, const FTSENT *p, int fd, const char *path)
 	if (fstat(fd, &sb) == -1)
 		goto bail;
 
+#ifdef __minix
+	/*
+	 * Skip the safety check on file systems where inodes are not static.
+	 * On such file systems, a file may legitimately be assigned a new
+	 * inode number due to being deleted and regenerated while we are
+	 * running.  This behavior is not POSIX compliant, but necessary for
+	 * certain types of file systems.  Currently, we assume that this
+	 * applies to all (and only) file systems that are not device backed.
+	 * In the future, we may have to obtain an appropriate flag through
+	 * statvfs(3) instead.
+	 */
+	if ((sb.st_ino != p->fts_ino || sb.st_dev != p->fts_dev)
+		&& major(sb.st_dev) != NONE_MAJOR) {
+#else
 	if (sb.st_ino != p->fts_ino || sb.st_dev != p->fts_dev) {
+#endif
 		errno = ENOENT;
 		goto bail;
 	}