These patches contain the following linux kernel bug fixes:
Description:
kernel: Memory detection problems with 32GB or more.
Symptom:
Kernel seems to hang if IPL'ed in an environment with 32GB or
more memory.
Problem:
The memory size is contained in a structure, which the kernel
accesses with the LH instruction. Since this instruction performs
sign extension this is a problem as soon as 32GB or more memory
is reported. In this case this will result in a very large value
loaded into a register which will be used in a loop, resulting in
a very long delay until the kernel finally continues operation.
If 64GB or more memory is defined, the kernel reads the memory
size from a wrong location, because of a wrong offset.
Solution:
Use different instruction that does not perform sign extension and
use correct offset.
Problem-ID:
15858
Description:
zfcp: Do port re-open only if test link fails.
Symptom:
Remote port is re-opened by zfcp module after an FC-2 problem
has being reported by Emulex adapter even if link is up.
Problem:
In case Emulex adapter reports an FC-2 problem zfcp module
re-opens a remote port without having checked the link status.
Solution:
Re-open remote port only if link is down.
Problem-ID:
16146
Description:
zfcp: Inproper handling of failed nameserver requests.
Symptom:
Ports are marked as failed if nameserver request
was not successfully sent.
Problem:
ZFCP_STATUS_PORT_INVALID_WWPN is set if nameserver request
could not be sent.
Solution:
Set ZFCP_STATUS_PORT_INVALID_WWPN only if response payload
for nameserver request indicates a reject of the request.
Problem-ID:
16120
Description:
zfcp: Problem in zfcp_fsf_send_els_handler() when D_ID assignment changes.
Symptom:
False port is re-opened if ADISC LS command failed on another port.
Problem:
D_ID assignment may be changed after ADISC command has been sent
and before ADISC response comes back.
Solution:
Save a pointer to zfcp_port structure as well as port's D_ID
when sending an ADISC.
Problem-ID:
16114
Description:
zfcp: Bug during adapter shutdown.
Symptom:
Message "zfcp: bug: There are still -1 FSF requests
pending on adapter 0.0.c500 after cleanup." occurred.
Problem:
When zfcp cleans up its fsf_req list, it is possible that the completion
for a member of that list is executed concurrently. This leads to the bug
message and finally to memory corruption, because memory is freed twice.
Solution:
Serialize access to fsf_req list between zfcp_fsf_req_dismiss_all
and zfcp_qdio_reqid_check and delete a fsf-request from the list
if its completion is triggered.
Problem-ID:
15658
Everybody should apply this patch.
To create the complete linux kernel sources, the following
patches need to be applied in sequence:
