Retransmission of lanplus request is broken. ipmi_lanplus_send_payload() keeps
current timeout in local variable 'timeout' and increases it by one with each
retransmission. But ipmi_lan_poll_recv() waits for response for session->timeout
seconds, which is not incremented -> the request is retransmitted only once,
when timeout and session->timeout equals. No other retransmissions are ever
sent.
In addition, the time is measured in seconds, therefore 'time(NULL) - ltime'
often equals 'timeout' and retransmission should be sent in this case (i.e.
'xmit' should be set).
Commit for Jan Safranek
When lan or lanplus session is being opened, ipmi_get_auth_capabilities_cmd() is
called twice. Now that we have -R and -N options it's not necessary to have such
weird code, especially with comment ``I'm not sure why we accept a failure for
the first call''.
Commit for Jan Safranek
Not all SEL entries have a "threshold value" specified, yet ipmitool sel elist
always displays a threshold value (as "0.0"), even when one is not specified.
Commit for Rob Swindell
Command "dcmi discover" fails with ``ERROR! This command is not available on this
platform'' unless the DCMI implementation advertises DCMI 1.0 or 1.1 conformance.
This means command will fail DCMI 1.5 capable hardware. Commit fixes this issue.
Commit for Rob Swindell
If any DCMI command fails, the function chk_rsp() in ipmi_dcmi.c displays the
incorrect byte as the completion code.
$ ipmitool dcmi power activate
DCMI request failed because: Unspecified error (dc)
The "dc" value printed is actually the "Group extension identification" byte
which *always* has the value 0xdc for DCMI responses.
Commit for Rob Swindell
The patch adds support for direct serial interface drivers (Terminal and Basic
mode).
Despite both drivers implement single and double bridging, it doesn't work at
the moment. Making this functionality to work depends on outcome of
ID#3611226(?).
Commit for Dmitry Bazhenov
Intel 82751 MAC has several deviations in its RMCP+ implementation while working
in the super pass-through mode.
This patch adds a OEM handle which allows IPMItool succesfully interract with
boards equipped with this NIC.
Commit for Dmitry Bazhenov
getpass() and getpassphrase() library functions return pointer to a static
data array. When free() is called on such pointer, it abnormally terminates
application.
Commit for Dmitry Bazhenov
Commit replaces strtol() calls with str2uchar() ones in order to mitigate
possible *flow via user input.
Also, limits are applied to 'Component ID' and 'Properties selector' now.
Reported-by: Dune Idaho
Commit replaces atoi() and strtoul() calls with wrappers in order to prevent
possible int *flow via user input.
Also, limits on user provided values are applied wherever possible.
Reported-by: Dune Idaho
Commit adds new function str2char() into 'lib/helper.c'. Also, function
is_fru_id() moves from 'lib/ipmi_fru.c' into 'lib/helper.c' in order to be
utilized by other modules.
Commit fixes bug where OS default values were over-ridden if the corresponding
enable/disable arguments weren't provided via command-line parameters.
Commit also changes IPMI shell default configuration setting from 'yes' to
'auto'.
Commit for Dmitry Bazhenov
Fixed a bug where connecting to a remote shelf manager via the lan
interface would always fail with:
``Activate Session error: Session sequence out of range Error: Unable
to establish LAN session''
This is due to a bug introduced in get_random() by fix for ID#3600963 where
braces were missing from an error check.
Commit for Dan Gora
ID: 3564701 - ipmitool 1.8.12 doesn't build on big endian architectures
Compilation would fail on big endian architecture because BSWAP_32 was not
defined in 'ipmi_chassis.c' and 'ipmi_pef.c'.
The problem was '#include <ipmitool/bswap.h>' depended on WORDS_BIGENDIAN which
comes from '<config.h>'. However, none of the previous header files
would include '<config.h>', so it remained undefined.
'bswap.h' gets included unconditionally now.
Commit for Dan Gora
