just discovered, trying to resurrect picobsd on -current, that the compiler in 6.x/7.x has become smart and, at least with the default compilation flags, will optimize out the "end_mfs_root" string from the object. As a consequence, any checks that the preloaded module does not overflow the allocated space (looking at the 'MFS Filesystem had better STOP here' string in the patched object) will fail. The attached patch fixes the issue. Any objection if i commit it ? cheers luigi Index: md.c =================================================================== RCS file: /prova/home/ncvs/src/sys/dev/md/md.c,v retrieving revision 1.159 diff -u -p -r1.159 md.c --- md.c 31 Oct 2005 15:41:19 -0000 1.159 +++ md.c 29 Jan 2006 12:57:22 -0000 _at__at_ -102,8 +102,13 _at__at_ SYSCTL_INT(_debug, OID_AUTO, mddebug, CT #if defined(MD_ROOT) && defined(MD_ROOT_SIZE) /* Image gets put here: */ -static u_char mfs_root[MD_ROOT_SIZE*1024] = "MFS Filesystem goes here"; -static u_char end_mfs_root[] __unused = "MFS Filesystem had better STOP here"; +static struct { + u_char start[MD_ROOT_SIZE*1024]; + u_char end[128]; +} mfs_root = { + .start = "MFS Filesystem goes here", + .end = "MFS Filesystem had better STOP here", +}; #endif static g_init_t g_md_init; _at__at_ -1139,7 +1144,7 _at__at_ g_md_init(struct g_class *mp __unused) g_topology_unlock(); #ifdef MD_ROOT_SIZE sx_xlock(&md_sx); - md_preloaded(mfs_root, MD_ROOT_SIZE * 1024); + md_preloaded(mfs_root.start, MD_ROOT_SIZE * 1024); sx_xunlock(&md_sx); #endif /* XXX: are preload_* static or do they need Giant ? */Received on Sun Jan 29 2006 - 12:04:42 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:51 UTC