On Sun Fire Servers, Using Dynamic Reconfiguration (DR) on System Boards With Permanent Memory May Fail Due to a Non Responsive QLC Driver

---

Impact

On Sun Fire servers, using Dynamic Reconfiguration (DR) on system boards with permanent memory may not complete due to a non responsive QLC driver.

Contributing Factors

This issue can occur in the following releases:

SPARC Platform

• Solaris 8 without patch 111097-14
• Solaris 9 without patch 113042-06

on the following platforms:

• Sun Fire Servers 3800/4800/4810/ 6800
• Sun Fire 12K/15K

The described issue only occurs under the following conditions:

1. On Host Bus Adaptors with QLC drivers installed.

2. During DR operations (detaching the system board using "cfgadm -c unconfigure" from the domain) on system boards with permanent memory.

To determine if a system is equipped with QLC drivers, run the following command:

    $ modinfo | grep qlc
    46 10291f33  25ff0 153   1  qlc (Qlogic FCA Driver v0.xxx)

To determine which system board hosts permanent memory, apply the following command as root:

    # cfgadm -val | grep permanent
    N0.SB0::memory   connected    configured   ok
    base address 0x0, 4194304 KBytes total, 720472 KBytes permanent

In the example above, SB0 is the system board hosting permanent memory.

Symptoms

Should the described issue occur, messages similar to the following are logged to the "/var/adm/messages" file:

    <date> <system> qlc: [ID 499440 kern.warning] WARNING: qlc(1) Fail suspend busy 1 flags 1184c
    <date> <system> sbdp: [ID 868600 kern.warning] WARNING: sbdp: failed to quiesce OS for copy-rename
    <date> <system> sbd: [ID 761800 kern.warning] WARNING: sbd_detach_memory: failed to move memory from board 0 to board

Note: A "non-zero" value will be seen when this issue occurs.

Workaround

To work around the described issue, do the following:

1. Find the qlc instance which is exhibiting the failure:

    # egrep "qlc.*Fail suspend busy" /var/adm/messages|tail -1
    Oct  2 16:54:07 v4u-4800d-doma qlc: [ID 499440 kern.warning] WARNING: qlc(1) 
    Fail suspend busy 1 flags 11844

In the above example, the qlc instance which is exhibiting the failure is "1"

2. Take the failing instance number and run grep "<instance> \"qlc\"" /etc/path_to_inst

    # grep "1 \"qlc\"" /etc/path_to_inst
    "/ssm@0,0/pci@1d,600000/SUNW,qlc@1" 1 "qlc"

3. Take the pci nexus (the text corresponding to "pci@1d,600000" in step 2) and run cfgadm -val|grep <pci nexus>

    # cfgadm -val|grep pci@1d,600000
    N0.IB8::pci1   connected   configured   ok   device /ssm@0,0/pci@1d,600000

4. Take the attachment point ID (in this example "N0.IB8::pci1") and run:

    # cfgadm -c unconfigure N0.IB8::pci1
    # cfgadm -c configure N0.IB8::pci1

Resolution

This issue is addressed in the following releases:

SPARC Platform

• Solaris 8 with patch 111097-14 or later
• Solaris 9 with patch 113042-06 or later

Modification History


Date: 18-MAR-2004

• State Resolved
• Updated Contributing Factors and Resolution sections

Attachments

This solution has no attachment