HP Support Information Digests

===============================================================================
o IT Resource Center World Wide Web Service
   ---------------------------------------------------

   If you subscribed through the IT Resource Center and would
   like to be REMOVED from this mailing list, access the
   IT Resource Center on the World Wide Web at:

     http://www.itresourcecenter.hp.com/

   Login using your IT Resource Center User ID and Password.
   Then select Support Information Digests (located under
   Maintenance and Support). You may then unsubscribe from the
   appropriate digest.

   To download a patch referenced below, access the
   IT Resource Center on the World Wide Web at:

     http://www.itresourcecenter.hp.com/

   Login using your IT Resource Center User ID and Password.
   Then select Individual Patches (under Maintenance and Support)
   to access the patch. You may also download a patch via anonymous
   ftp(1) from ftp.itrc.hp.com.
===============================================================================


Digest Name: weekly HP-UX series 700 10.X patch digest
    Created: Sun Jun 23 3:05:09 PDT 2002

Table of Contents:

Document ID Title
--------------- -----------
PHKL_26980 s700 10.20 VxFS (JFS) mount, fsck, vx_real_iget deadlock

The documents are listed below.
-------------------------------------------------------------------------------


Document ID: PHKL_26980
Date Loaded: 20020617
      Title: s700 10.20 VxFS (JFS) mount, fsck, vx_real_iget deadlock

Patch Name: PHKL_26980

Patch Description: s700 10.20 VxFS (JFS) mount, fsck, vx_real_iget deadlock

Creation Date: 02/05/16

Post Date: 02/06/17

Hardware Platforms - OS Releases:
        s700: 10.20

Products: N/A

Filesets:
        JournalFS.VXFS-BASE-KRN

Automatic Reboot?: Yes

Status: General Release

Critical:
        Yes
        PHKL_26980: HANG
        PHKL_18197: CORRUPTION
        PHKL_17716: PANIC
        PHKL_12339: HANG
        PHKL_12007: CORRUPTION

Path Name: /hp-ux_patches/s700/10.X/PHKL_26980

Symptoms:
        PHKL_26980:
        (SR: 8606214705 CR: JAGad83897)
        Deadlock of processes waiting for vxfs inodes. There
        is a situation in vx_real_iget where such a deadlock can
        occur with process 1 waiting for inode A (while holding
        inode B locked) while process 2 is holding inode A and
        waiting for inode B. One symptom to look for is hung
        process sleeping in vx_real_iget(). The stack trace may
        look similar to

        _swtch+0x138
        real_sleep+0x234
        vx_sleep_lock+0xa8
        vx_ilock+0x14
        vx_real_iget+0x184
        vx_iget+0x3c
        ......

        PHKL_21594:
        (SR: 8606135341 CR: JAGad04475)
        Inactive inodes are not getting cleared after 110 days.

        (SR: 8606104878 CR: JAGab72639)
        System performance is unacceptable (slows down)
        due to ibmaplock contention even when inode
        is in memory.

        PHKL_19539:
        ( SR: 1653302711 DTS: JAGab15428 )
        A VxFS file system cannot be mounted if it has more than 8
        million inodes.

        PHKL_18197:
        ( SR: 1653297531 DTS: )
        On vxfs filesystems where a particular type of inode
        corruption occurs, all future mounts of that filesystem
        fail. In the message buf will be the message:
            vx_iget - [fs name] file system invalid inode number [#]
        This message will possibly be repeated many times. Future
        attempts to mount this file system will fail with either:
            vxfs mount: [device] is not a vxfs file system
         or vxfs mount: [device] is corrupted. needs checking
        At that point, if an fsck is performed the file system can
        be mounted, but further use of the file system may cause
        the vx_iget message to occur again, and the file system to
        again be unmountable.

        PHKL_17716:
        ( SR: 4701411298 DTS: JAGaa45918 )
        A panic may occur in vx_spinlock() when trying to mount
        a corrupt VxFS file system.

        This patch provides a fix so that a failure is reported
        in this case, but a panic will not occur.

        PHKL_12339:
        ( SR: 5003383307 DTS: JAGaa00693 )
        Mounting a VxFS Version 2 filesystem containing a large
        amount of files (in excess of 800,000) can take several
        minutes to complete. During the time it takes to mount
        the filesystem the system appears to be hung.

        This patch provides improved kernel routines which
        eliminate this performance problem.

        PHKL_12007:
        ( SR: 4701361758 DTS: DSDe438092 )
        10.20 fsck on vxfs (version 3) filesystem fails with
         the following error:
          # fsck -o full -y /dev/vg01/lvol3
          pass0 - checking structural files
          pass1 - checking inode sanity and blocks
          pass2 - checking directory linkage
          pass3 - checking reference counts
          pass4 - checking resource maps
          small extent in iau inode
        This occured after a normal shutdown. The filesystem
        is unusable and must be newfs'd.
        This patch provides the kernel half of a solution for an
        fsck defect. The commands patch PHCO_11908 must also be
        applied to install the full solution. Without PHCO_11908
        installed, this patch will have no impact.

Defect Description:
        PHKL_26980:
        (SR: 8606214705 CR: JAGad83897)
        In vx_real_iget() an inode is kept locked, while waiting
        to lock another inode. The inode which is kept locked is
        not needed and is unlocked when the other inode has been
        acquired. Another process trying for these inodes in
        reverse order will cause the deadlock.

        Resolution:
        The inode which is not needed is released before trying
        for another inode.

        PHKL_21594:
        (SR: 8606135341 CR: JAGad04475)
        Prior to 110 days inactive inodes are removed early from the
        inode cache. After 110 days inactive inodes are never
        removed from the inode cache. The JFS inode cache or inode
        table is dynamic it keeps expanding and contracting, but
        after the system is up for over 110 days the cache does not
        reduce in size, it grows to the max and stays the same after
        that. Customer monitored the inode size using:
        # sar -v 1 1
        16:29:09 text-sz ov proc-sz ov inod-sz ov file-sz ov
        16:29:10 N/A N/A 272/1620 0 3500/3500 0 1137/6010 0

        Resolution:
        The variable (i_ftime) which is assigned the time inode is
        put on free list is incorrectly set to lbolt, it should be
        set to vx_lbolt (system time) in the function vx_inactive()
        which inactivates an inode.

        (SR: 8606104878 CR: JAGab72639)
        The vx_iget() function grabs ibmaplocks exclusively every
        time it is called, even when the inode is in memory. This
        is slowing down the system.

        Resolution:
        Changes vx_iget() function to obtain ibmaplock, only
        if inode is not in memory.

        PHKL_19539:
        ( SR: 1653302711 DTS: JAGab15428 )
        To reproduce the problem, create a VxFS file system with
        more than 8 million files. The file system cannot be
        mounted without this patch.

        Resolution:
        The problem occurs because a 32 bit integer overflows when
        the number of inodes reaches 8 million. The variable has
        been changed to a 64 bit integer.

        PHKL_18197:
        ( SR: 1653297531 DTS: )
        The vxfs kernel was not properly handling sparse IFILT
        files. This was due to fsh_ninode being incorrectly set in
        this case, as well as the end of the IFILT list being
        calculated incorrectly.

        Resolution:
        vx_fsetialloc() was updated to properly update fsh_ninode
        in the case of a sparse IFILT file. vx_olt_ilistadd() was
        updated to properly find the end of the IFILT list in this
        case. This prevents a corrupt filesystem from getting
        worse, and will not fix the inital cause of the corruption.

        PHKL_17716:
        ( SR: 4701411298 DTS: JAGaa45918 )
        SR: 4701411298 DTS: JAGaa45918
        The problem can be reproduced by using fsdb to corrupt inode
        2 (the root inode) of the file system.

        Resolution:
        The problem has been corrected so that the mount will fail
        instead of causing a panic if the file system is corrupted
        in this manner since a corrupted file system should not be
        mounted.

        PHKL_12339:
        ( SR: 5003383307 DTS: JAGaa00693 )
        Mounting a version 2 VxFS filesystem containing many
        (>800K) files takes several minutes.

        PHKL_12007:
        ( SR: 4701361758 DTS: DSDe438092 )
        This is the kernel part of the solution to that fsck problem
        that PHCO_11908 addresses. It involves a VxFS filesystem.

SR:
        1653297531 1653302711 4701361758 4701411298 5003383307
        8606104878 8606135341 8606214705

Patch Files:
        /usr/conf/lib/libvxfs_base.a(vx_fsetsubr.o)
        /usr/conf/lib/libvxfs_base.a(vx_inode.o)
        /usr/conf/lib/libvxfs_base.a(vx_oltmount.o)

what(1) Output:
        /usr/conf/lib/libvxfs_base.a(vx_fsetsubr.o):
                vx_fsetsubr.c $Date: 99/04/09 10:15:28 $ $Revision:
                        1.2.98.15 $ PATCH_10.20 (PHKL_18197)
        /usr/conf/lib/libvxfs_base.a(vx_inode.o):
                vx_inode.c $Date: 2002/05/02 01:05:49 $ $Revision: 1
                        .7.98.23 $ PATCH_10.20 (PHKL_26980)
        /usr/conf/lib/libvxfs_base.a(vx_oltmount.o):
                vx_oltmount.c $Date: 99/08/20 07:01:32 $ $Revision:
                        1.6.98.17 $ PATCH_10.20 (PHKL_19539)

cksum(1) Output:
        3172264055 21208 /usr/conf/lib/libvxfs_base.a(vx_fsetsubr.o)
        3695379388 45024 /usr/conf/lib/libvxfs_base.a(vx_inode.o)
        3177947342 27968 /usr/conf/lib/libvxfs_base.a(vx_oltmount.o)

Patch Conflicts: None

Patch Dependencies:
        s700: 10.20: PHKL_16750

Hardware Dependencies: None

Other Dependencies: None

Supersedes:
        PHKL_12007 PHKL_12339 PHKL_17716 PHKL_18197 PHKL_19539 PHKL_21594

Equivalent Patches:
        PHKL_26981:
        s800: 10.20

Patch Package Size: 160 KBytes

Installation Instructions:
        Please review all instructions and the Hewlett-Packard
        SupportLine User Guide or your Hewlett-Packard support terms
        and conditions for precautions, scope of license,
        restrictions, and, limitation of liability and warranties,
        before installing this patch.
        ------------------------------------------------------------
        1. Back up your system before installing a patch.

        2. Login as root.

        3. Copy the patch to the /tmp directory.

        4. Move to the /tmp directory and unshar the patch:

                cd /tmp
                sh PHKL_26980

        5a. For a standalone system, run swinstall to install the
            patch:

                swinstall -x autoreboot=true -x match_target=true \
                        -s /tmp/PHKL_26980.depot

        By default swinstall will archive the original software in
        /var/adm/sw/patch/PHKL_26980. If you do not wish to retain a
        copy of the original software, you can create an empty file
        named /var/adm/sw/patch/PATCH_NOSAVE.

        WARNING: If this file exists when a patch is installed, the
                 patch cannot be deinstalled. Please be careful
                 when using this feature.

        It is recommended that you move the PHKL_26980.text file to
        /var/adm/sw/patch for future reference.

        To put this patch on a magnetic tape and install from the
        tape drive, use the command:

                dd if=/tmp/PHKL_26980.depot of=/dev/rmt/0m bs=2k

Special Installation Instructions:
        The kernel patch PHKL_12007 provides the kernel half of a
        solution for an fsck defect. The commands patch PHCO_11908
        must also be applied to achieve the full solution. Without
        PHCO_11908 installed, this patch will have no impact.

        This patch depends on base patch PHKL_16750.
        For successful installation, please ensure that PHKL_16750
        is in the same depot with this patch, or PHKL_16750 is
        already installed.
-----End of Document ID: PHKL_26980------------------------------------------

• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]