PROBLEM: (57336, 90066, BCGM51RKR) (PATCH ID: OSF520-223) ******** This patch fixes two issues: - If multiple processes attempted to access the same file at the same time and access to the file should have been allowed by an ACL on the file, access may have been denied instead. Now access will be allowed as expected. - If the ACL on a file was corrupted the corrupted data was being passed into the kernel causing a variety of problems, including panics in kernel malloc/free after it was called by the ACL code. Now more of the ACL data is being validated. The system will respond to the corrupted data with an "Invalid Argument" error. One example panic: trap: invalid memory write access from kernel mode 3 _XentMM 4 free_common 5 free 6 kfree 7 sp_delete_ir 8 acl_ir_cache_delete 9 LOCK_DONE_SECATTR PROBLEM: (78435) (PATCH ID: OSF520-091) ******** When an error is seen during an AutoFS auto-mount attempt involving a direct map entry, further uses of the same intercept point will fail to auto-mount as expected. PROBLEM: (90713, 91327) (PATCH ID: OSF520-290) ******** Using vdump to illustate the first two problems, if "/foo/bar" was an indirect map entry intercept, executing "vdump -D /foo/bar" would result in an error vdump: error accessing file system ; [2] No such file or directory since an auto-mount would not be induced. After the auto-mount was induced, execution of the command would yield: vdump: error reading extend attributes for directory <.> The typical symptom of the final problem is a hang of user-space code which trips an AutoFS intercept point that happens to be served locally. Again, this will only happen in non-clustered environments. PROBLEM: (84283) (PATCH ID: OSF520-027) ******** When abort_prog() is called by msg_snd() in msgqueue_util.c, the printf used requires three arguments. Only two were given. This would cause a core dump. The missing argument has been added. The problem was found by code inspection while working on an internationalization of messages qar. PROBLEM: (HPAQ21K51, HPAQ412CK) (PATCH ID: OSF520-034) ******** This patch fixes the following problems in the following commands: vdump: - failed to flag compressed extended attributes records that are split across a vdump BLOCK boundary - Correct "Rewinding" message to avoid a segfault with Internationalized messages. vrestore: - fails to properly handle extended attributes records in compressed archives. This results in malloc failures, proplist corruption, program abort, program crashes due to segfault or invalid memory access, and the display of the error message "error setting extended attributes" - fails to set extended attributes due to confusion over selective restore of the file or directory associated. Also results in display of error message "error setting extended attributes". - selective restore of hardlinked files is incomplete when they exist in different directories (fails to create directory for second occurrence of file with same inode number) PROBLEM: (88898) (PATCH ID: OSF520-273) ******** This patch eliminates what are ultimately futile mount attempts by autofsd when the top-level directory of a direct hierarchical map entry cannot be auto-mounted. PROBLEM: (90772) (PATCH ID: OSF520-270) ******** The problem fixed by this patch occurs when an auto-mount entry with replicated servers contains mount options, and is written in the style of : "key -options host1,host2:/dir" . If host2 is selected as the server, the specified mount options will not be used. PROBLEM: (BCSM31S18, 79413) (PATCH ID: OSF520-005) ******** This patch fixes a problem where the tar -F (Fasttar) option ignores files named "err" but doesn't ignore files named "errs" and directories named "SCCS" and "RCS". PROBLEM: (DM07739) (PATCH ID: OSF520-037) ******** This patch corrects pax/tar/cpio to properly extract explicitly specified files. When an archive contained a file with extended attributes and a different file (occurring later in the archive) was specified to be extracted, improper buffer pointer management resulted in the following display (example uses tar): tar: /dev/nrmt0h : This doesn't look like a tar archive tar: /dev/nrmt0h : Skipping to next file... tar: Memory allocation failed for extended data while reading : Not enough space The directory option was similarly affected. In this case the information for the specified file was not reported. PROBLEM: (90713) (PATCH ID: OSF520-372) ******** This patch ensures consistant behaviors by commands when walking the file namespace in the presence of AutoFS. Locally served auto-mounted file systems, which are implemented via symbolic links, are now dealt with by these commands in the same fashion as remotely served auto-mounted file systems. PROBLEM: (92363) (PATCH ID: OSF520-412) ******** This patch adds support to both AutoFS and Automount, for Linux /etc/exports entries containing wildcards in host names. Previously, auto-mount attempts for such entries would fail. PROBLEM: (GB_G03885, DE_G03607, CA1Q70856, 89383, 89386) (PATCH ID: OSF520-456) ******** This patch fixes a problem where aufofsd and autofsmount daemon do not properly parse "*" in map files. "*" entries in NIS map files will not be expanded properly. PROBLEM: (94406) (PATCH ID: OSF520-635) ******** With ACLs enabled, non-root users may be denied access to autofs indirect and special mounted filesystems. Non-root users may also be denied access to other filesystems that don't support ACLs (i.e. cdfs) as well. This patch corrects this problem. PROBLEM: (94940) (PATCH ID: OSF520-776) ******** This patch fixes autofsd so that it will correctly account for symbolic links in the path specified by direct map entry keys, when attempting an unmount. Without this fix, any auto-mounted filesystems associated with this type of map entry will never get auto-unmounted. PROBLEM: (KAOB83087) (PATCH ID: OSF520-739) ******** This patch fixes a one byte gap/hole in the maximum size in the tar command before an extended header record is used (8589934591 (octal 77777777777)). PROBLEM: (94011) (PATCH ID: OSF520-633) ******** This patch fixes AutoFS auto-mount failures due to the early termination of auto-mount inducing lookups and results in errors returned to operations generating such lookups. The early termination results from the receipt of an unexpected signal while the kernel is waiting a response from the autofsd daemon. The scenario that exposed this defect involved SIGCHLD in which a parent process spawned a child, caused an auto-mount, and the child exited while the kernel was waiting for the auto-mount to complete. This caused the auto-mount to fail. Since this is timing-sensitive, one may or may not see this problem reproduce easily. PROBLEM: (117-1-20865) (PATCH ID: OSF520-617) ******** This patch corrects a problem in which "find -ls" displayed an incorrect number of blocks, due to a wraparound to zero of the value in this data field. PROBLEM: (92789) (PATCH ID: OSF520-575) ******** This patch is in support of a Cluster patch to reduce DLM lock contention by AutoFS. PROBLEM: (93840) (PATCH ID: OSF520-685) ******** The defect being fixed by this patch may result in multiply mounted, AutoFS auto-mounted, NFS file systems. This results from a race condition that may occur when an auto-mount request associated with an indirect map entry is interrupted. The filesystem in question can still be accessed as normal. Auto-Unmounts will, however, fail. Such failures will be logged in the daemon.log . PROBLEM: (94501) (PATCH ID: OSF520-643) ******** This patch corrects a defect wherein auto-mount map files provided via NIS, and included via a "+" directive in a file, may not be processed by autofsmount. The resulting symptom would be missing intercept mount points and failing auto-mounts for those cases. PROBLEM: (93963, 93598) (PATCH ID: OSF520-582) ******** If an auto-mount map entry specifies multiple possible sites from which to mount a remote filesystem, the current implementation may contact the same system's mount daemon multiple times when checking for its responsiveness. This patch eliminates such duplicate checks. PROBLEM: (93298) (PATCH ID: OSF520-509) ******** This patch corrects a defect in autofsd wherein NIS-provided auto-mount map entries using the wildcard * key are not recognized. The problem will be seen as a failure to auto-mount any file systems relying upon such a map entry. PROBLEM: (94823) (PATCH ID: OSF520-753) ******** This patch corrects two diagnostic messages so that correct data is printed or logged. PROBLEM: (94052) (PATCH ID: OSF520-598) ******** The problem fixed by this patch may lead to the timeout failure of auto-mount requests. When the problem occurs, a large number of autofsd processes will exist, and will be waiting for a chance to process their request. Their kernel stack will indicate they are sleeping for a lock within a CFS routine. PROBLEM: (74390, 83663, 93668, 94171, 94771) (PATCH ID: OSF520-955) ******** * A file/directory name is displayed along with the error message when vrestore command fails to set a property list * Vrestore does not dump core when when a tape has a smaller blocksize than expected * Vrestore handles no-rewind tapes properly. * Vrestore reads environment variable for user defined device name. * Vrestore sets the attributes to top level directory if user desires. PROBLEM: (95084) (PATCH ID: OSF520-821) ******** When the input to vrestore is from a remote system through a pipe, as in: % rsh remotemchine cat filearchive | vrestore -tf - vrestore *may* sometimes fail if the very first read from the pipe fails to provide the complete number of bytes requested in the read(). This patch fixes that problem. PROBLEM: (91296, 76704, 78982, 79298) (PATCH ID: OSF520-913) ******** This submit request fixes the following problems. 1) If vdump or vrestore are interrupted they will dump core. The expected behavior is to ask the user to confirm if the program should be aborted. 2) In some locales, vdump and vrestore programs displays some incorrect messages. This has been corrected now. 3) vrestore with -l option displays the saveset structure. Now this option has been enhanced to show bitfile attributes on files. Bit files attributes are usually set on files in advfs fileset using 'usr/sbin/chfile'. 4) vdump program makes some unneccessary calls to backup property list information, which consumes time. Now this has been corrected in vdump to make these calls only when required. PROBLEM: (95642) (PATCH ID: OSF520-1084) ******** This patch reports errors emerging from the failure of the close() call for devices/files which are on an NFS mount point. PROBLEM: (92731) (PATCH ID: OSF520-863) ******** Archive programs vdump and vrestore are modified to allow the user to request larger record sizes. Previously the maximum was 64k, now we allow 2048k.