From IPMI Spec, Chapter 22.22 Set Channel Access Command
Table 22, Set Channel Access Command
Byte#2, Bit#5 is "PEF Alerting Enable/Disable"
And the bit value:
0b = enable PEF Alerting
1b = disable PEF Alerting on this channel
In current code, alert "on" set Bit#5 to 1 and alert "off" set Bit#5 to
0, it's straightforward but just opposite of IPMI spec bit definition.
Resolvesipmitool/ipmitool#247
Reported-by: Ryan Fang <Ryan.Fang@quantatw.com>
Signed-off-by: Alexander Amelkin <alexander@amelkin.msk.ru>
Previously, ipmitool would fail to run if the local copy of the IANA PEN
registry could not be parsed.
When the registry is not available the manufacturer will be "Unknown" but
ipmitool will otherwise function so should not be considered fatal.
Also, fix an issue with improperly handling the `oem_info_list_load`
return value. Previously, in `ipmi_oem_info_init`, if `oem_info_list_load`
returned a negative value due to the registry file not existing, an
improper count would cause `oem_info_init_from_list` to aallocate a list
that didn't encompass the full header/tail list.
IANA PEN registry open failed: No such file or directory
Allocating 3 entries
[ 1] 16777214 | A Debug Assisting Company, Ltd.
[ 0] 1048575 | Unspecified
Now, use a signed int and ensure a valid count of loaded OEMs is used.
Signed-off-by: Vincent Fazio <vfazio@gmail.com>
For computers using the arm64 of the Apple chip, the link requires strict
alignment of pointers in 32-bit form. Replace the struct vFlashstr to valstr.
Replace the Function get_vFlash_compcode_str to val2str.
Resolvesipmitool/ipmitool#332
Signed-off-by: SliderLu <314238828@qq.com>
* Add a decoder for System Management records in
the Multirecord area
* Refactor GUID/UUID decoding: Use the same code for `mc guid`
and for `fru print` to decode the GUID and System Unique ID
in System Management records in the Multirecord Area.
* Fix some type errors in calls to printf/sprintf in GUID decoder
Signed-off-by: Alexander Amelkin <alexander@amelkin.msk.ru>
Building ipmitool with UBSAN and I got:
```
...
xxx/ipmitool/include/ipmitool/helper.h:191:14: runtime error: left shift of 160
by 24 places cannot be represented in type 'int'
#0 0x55bddaa56f11 in ipmi32toh
xxx/ipmitool/include/ipmitool/helper.h:191:14
```
Tested: with this, I tested ipmitool again and the issue disappeared.
Resolved: ipmitool/ipmitool#352
Signed-off-by: Tom Tung <shes050117@gmail.com>
Add 'soft reset' and 'power-up by RTC wakeup' causes
from IPMI 2.0 spec.
Resolvesipmitool/ipmitool#329
Signed-off-by: Alexander Amelkin <alexander@amelkin.msk.ru>
The IPM_DEV_FWREV1_MAJOR_MASK was incorrectly defined as 0x3F.
The spec indicates that the major firmware revision has 0~6 valid bits,
so the mask shall be 0x7F.
Tested: Verify a BMC that has a major version that is large than 0x3F,
and ipmitool shows the correct major version.
Signed-off-by: Lei YU <yulei.sh@bytedance.com>
There was a wrong offset for the "Power Supply Inactive" flag,
hence the trigger on this event type could not be set.
Resolvesipmitool/ipmitool#280
Signed-off-by: Alexander Amelkin <alexander@amelkin.msk.ru>
Some of the header guard names in ipmitool duplicated with the ones in
freeipmi.
* ipmitool/ipmi_fru.h and freeipmi/fru/ipmi-fru.h both uses IPMI_FRU_H
* ipmitool/ipmi_sdr.h and freeipmi/sdr/ipmi-sdr.h both uses IPMI_SDR_H
* ipmitool/ipmi_sel.h and freeipmi/sel/ipmi-sel.h both uses IPMI_SEL_H
This is problematic as including the 1st will prevent the 2nd from being
included, leading to the loss of any declarations, inline definitions,
or other "#includes" in the 2nd header.
For example, including ipmitool/ipmi_sel.h and freeipmi/freeipmi.h fails
to build.
$ cat test.c
#include <ipmitool/ipmi_sel.h>
#include <freeipmi/freeipmi.h>
$ gcc test.c -I${HOME}/src/oss/ipmitool/include
In file included from /usr/include/freeipmi/freeipmi.h:90,
from test.c:2:
/usr/include/freeipmi/sdr/oem/ipmi-sdr-oem-intel-node-manager.h:48:44: error: unknown type name 'ipmi_sdr_ctx_t'; did you mean 'ipmi_fru_ctx_t'?
48 | int ipmi_sdr_oem_parse_intel_node_manager (ipmi_sdr_ctx_t ctx,
| ^~~~~~~~~~~~~~
| ipmi_fru_ctx_t
Remove header guards and use "#pragma once" instead. In this way, we
don't have to do manual management to avoid name clashes.
* src/plugins/lan/md5.h is left as is being an external header
originally.
* src/plugins/imb/imbapi.h is a convoluted header consisting of multiple
header guards. Let's just add "#pragma once" and leave header guards
as is for now.
Signed-off-by: Bing-Hua Wang <binghuawang@ami.com>
The following commands is used on Linux to convert line endings to LF.
git config core.autocrlf input
git add --renormalize .
git config --unset core.autocrlf
It turns out there was only ipmi_kontronoem.h not using LF.
Signed-off-by: Bing-Hua Wang <binghuawang@ami.com>
- Drop a useless `extern` definition
- Remove static variable `errorCount` from header, move it to
the c file where it is actually used if needed
Related to ipmitool/ipmitool#220
Signed-off-by: Alexander Amelkin <alexander@amelkin.msk.ru>
Reduce code duplication by extracting option names, types,
and value ranges into a separate structure, and rewriting
the option parsing code without mixing it with the data.
Signed-off-by: Alexander Amelkin <alexander@amelkin.msk.ru>
ipmi_intf.h was missing a couple of declarations for the functions
used by fru, sdr and hpmfwupg modules. Add those declarations
to ipmi_intf.h and remove local declarations.
This fixes a couple of compilation warnings.
Signed-off-by: Alexander Amelkin <alexander@amelkin.msk.ru>
IPMI 2.0 specification is quite inconsistent about system interfaces.
They have section 1.7.16 "System Interfaces" that clearly states that
there are FOUR system interfaces (KCS, SMIC, BT and SSIF), but then they
have section 1.7.31 saying that "It is mandatory to implement a system
interface that is compatible with one of the **three** specified system
interfaces" without specifying which three of the four interfaces are
meant. Then in section 6 "IPMI Messaging interfaces" they again say that
"As mentioned earlier, there are three System Interface implementations
specified for the BMC: SMIC, KCS, and BT". Is all looks like during
update from 1.5 to 2.0 they have updated section 1.7.16, but forgot to
update Table 6-3, section 1.7.31 and section 6. Yet again, there is 'Get
System Interface Capabilities' command that has a parameter 'System
Interface Type' that can specify that SI is of SSIF type.
All that have lead to a situation where some BMC manufacturers treated
the specification as if it prohibited specifying media type 0xC
(which is "System Interface") for system interfaces using SSIF
(SMBus Sustem Interface), and so they specified an SMBUS media type
for their system interface channels.
As a result, ipmitool failed to properly send event data via such
system interfaces as it treated them as non-system and didn't add
the required Generator ID.
To mitigate the inconsistency of IPMI specification and yet not
ask BMC manufacturers to alter their code, thus increasing compatibility
with legacy BMCs, this commit adds checking of current interface number.
The system interface, according to Table 6-1 of IPMI Specification is
required to have channel number 15 (0Fh). So with this commit the
generator ID is added for any interfaces that are either marked
as media type 0Ch 'System Interface' or have channel number 0Fh.
Resolvesipmitool/ipmitool#111
Signed-off-by: Alexander Amelkin <alexander@amelkin.msk.ru>
If a manufacturer's IANA PEN (aka manufacturer ID) was above
65535, it wasn't reported properly. Luckily there are no such
IDs so far, the biggest is 54077 as of 2019/06/18.
There is, however, an ID 0xFFFFFE used by fake_ipmistack
for debug purposes, and it was not reported correctly.
This commit expands the value argument to string searching functions
from 16-bit to 32-bit to allow for any possible IANA PEN.
Fixes: 73d6af57827fc85e78c700ca1dff00b3dbc63948
Signed-off-by: Alexander Amelkin <alexander@amelkin.msk.ru>
Previously, the OEM names dictionary was compiled in and
updating it required rebuilding of `ipmitool`, thus taking a
long time for newly registered OEMs to get supported by the tool.
Building also required a direct internet connection to succeed.
With this commit, the OEM enterprise dictionary is now loaded from
either ${HOME}/.local/usr/share/misc/enterprise-numbers or from
/usr/share/misc/enterprise-numbers (in that precedence).
Those files can be downloaded from iana.org at
http://www.iana.org/assignments/enterprise-numbers
Partially resolvesipmitool/ipmitool#11
Fixes: 9d41136c9b7c7d392f1a3f3adeb6d7fe3bd3135e
Signed-off-by: Alexander Amelkin <alexander@amelkin.msk.ru>
Renamed the oem strings containing Newisys to Viking Enterprise Solutions.
IANA 9237
Built and tested with AMI's IPMI stack LTS 12.01.
Resolvesipmitool/ipmitool#124
Signed-off-by: dan mcgee <dan.mcgee@vikingenterprise.com>
Signed-off-by: dan mcgee <dan.mcgee@sanmina.com>
Some commands may return command-specific completion codes.
Now they are all reported as 'Unknown'.
Add helper functions to support such command-specific codes.
Command handlers will need to define their own valstr arrays
with completion code descriptions and then use specific_val2str()
instead of generic val2str() to convert the completion code into
a string.
Also reduce code duplication in helper.c
Signed-off-by: Alexander Amelkin <alexander@amelkin.msk.ru>
The function converts a set of command line arguments representing
byte values into a byte buffer and verifies each individual value
to be a valid data byte.
Signed-off-by: Alexander Amelkin <alexander@amelkin.msk.ru>
The program would crash if the BMC returned an out of range (>90)
unit type for a full sensor record. This commit adds a range check
and also add support for IPMI 2.0 additional unit types 91 and 92
("fatal error" and "grams").
Resolvesipmitool/ipmitool#118
Signed-off-by: Alexander Amelkin <alexander@amelkin.msk.ru>
Make the argument to free_n() compatible with any pointers,
thus reducing the number of compilation warnings.
End-user-impact: None
Signed-off-by: Alexander Amelkin <alexander@amelkin.msk.ru>
Fixup the following array bounds checking bugs:
[lib/ipmi_fru.c:1003]: (style) Array index 'i' is
used before limits check.
[lib/ipmi_fru.c:1127]: (style) Array index 'i' is
used before limits check.
[lib/ipmi_fru.c:1262]: (style) Array index 'i' is
used before limits check.
Signed-off-by: Patrick Venture <venture@google.com>
free_n() will free the memory and clear the pointer, which will reduce
the probability a developer will forget to clear the pointer after
freeing.
Resolves: #79
Signed-off-by: Patrick Venture <venture@google.com>
Move led color static array from header to ipmi_picmg module
and introduce `picmg_led_color_str()` function for use in
ipmi_vita module.
Partially resolvesipmitool/ipmitool#13
Signed-off-by: Alexander Amelkin <alexander@amelkin.msk.ru>
Move static objects declared in headers to the source files where
they're used.
Partially resolvesipmitool/ipmitool#13
Signed-off-by: Patrick Venture <venture@google.com>
Cleanup all unused-parameter warnings. Each warning was examined to
verify it wasn't simply a case of a build macro difference.
Partially resolvesipmitool/ipmitool#13
Signed-off-by: Patrick Venture <venture@google.com>
Use __UNUSED__ macro defined in helper.h instead of the gcc-specific
attribute unused macro.
Partially resolvesipmitool/ipmitool#13
Signed-off-by: Patrick Venture <venture@google.com>
FRU board mfg. date uses a different value for 'unspecified'
timestamp than the general IPMI specification.
This commit makes ekanalyzer and fru commands process unspecified
FRU dates properly, displaying 'Unspecified' instead of
'Mon Jan 1 03:00:00 1996'.
Resolvesipmitool/ipmitool#57
Signed-off-by: Alexander Amelkin <alexander@amelkin.msk.ru>
Several functions were passed a struct ipmi_intf* even though it was
unused. This removes the unused parameters and changes calls and
signatures.
Partially resolvesipmitool/ipmitool#13
Signed-off-by: Vernon Mauery <vernon.mauery@intel.com>
Before this fix `ipmitool` would complain that VLAN ID
is out of range when trying to disable an already disabled VLAN
on a lan channel. With this fix it will properly report that
VLAN is already disabled.
Resolvesipmitool/ipmitool#55
Signed-off-by: Alexander Amelkin <alexander@amelkin.msk.ru>
There is a bug in gcc since 4.3.2 and still not fixed in 8.1.0.
Even if __attribute__((format(strftime... is specified for a
wrapper function around strftime, gcc still complains about strftime
being called from the wrapper with a "non-literal" format argument.
See https://gcc.gnu.org/bugzilla/show_bug.cgi?id=39438
This commit adds 'ugly hacks' from that discussion to call strftime()
from strftime-formatted wrappers and silence the warnings.
Partially resolvesipmitool/ipmitool#23
Signed-off-by: Alexander Amelkin <alexander@amelkin.msk.ru>
Handle all date/time stamps uniformly, taking in account the host
endianness. Respect the local time zone and the '-Z' option for
all ipmitool commands. Unify the date and time formatting.
Add correct handling of IPMI timestamps 'since system startup'
and for 'unspecified' timestamps.
Partially resolvesipmitool/ipmitool#23
Signed-off-by: Alexander Amelkin <alexander@amelkin.msk.ru>
Some minor formatting corrections.
Also introduced a new helper function to reduce nesting level.
Signed-off-by: Alexander Amelkin <alexander@amelkin.msk.ru>
Based on current crypto alogrithms, one could rank cipher suites along
these lines:
17 > 3 >> all the rest
17 and 3 are the only cipher suites that implement any sort of
confidentiality alogorithm that is secure. In addition, any hmac-md5 or
md5 integrity algorithm used in integrity is weak at best and dangerous
for authentication.
This could possibly be enabled in a simpler mechanism by simply checking
for 17 and then choosing it before falling back to 3, but the way this
is implemented, it makes it easy to change the list of acceptable
algorithms from two to three or more items.
Resolvesipmitool/ipmitool#29
Signed-off-by: Vernon Mauery <vernon.mauery@intel.com>
With this commit the GUID encoding is now by default detected
automatically based on the validity of the version field,
and the timestamp (for time-based version 1 GUIDs).
The version is considered valid if it is 1 through 5.
The timestamp is considered valid if the year is past UNIX Epoch
and before the current year.
Resolvesipmitool/ipmitool#25
Signed-off-by: Alexander Amelkin <alexander@amelkin.msk.ru>
Before this commit the 'Timestamp' line was always printed
for all versions of GUID, even for non-time-based ones.
Plus, only the time_low field was used, and it was used as if
it contained seconds since UNIX Epoch, which it didn't. In fact
this field along with other time_* fields constitute a single
60-bit value representing the count of 100ns intervals since
adoption of Gregorial calendar (00:00:00.00 15 Oct 1582).
For non-time-based versions of GUID, the time_* fields do
not represent any time at all.
So, after this commit, the timestamp will be properly decoded
for time-based GUID version 1 only. For other versions the
'Timestamp' line will not be displayed. A line showing the
GUID version will be added to the output.
Partially resolvesipmitool/ipmitool#25
Signed-off-by: Alexander Amelkin <alexander@amelkin.msk.ru>
There are lots of BMC implementations that violate the IPMI
specification in regard to GUID encoding and instead encode
GUIDs according to either RFC4122 or SMBIOS specifications.
This commit restores the default behavior of `mc guid` to
SMBIOS-based decoding and adds options to allow for decoding
according to IPMI or RFC4122 specifications.
It also allows to simply dump the received GUID as is
without any parsing.
It is possible that in future versions `ipmitool` will
change default behavior to 'ipmi' instead of 'smbios'.
Partially resolvesipmitool/ipmitool#25
Signed-off-by: Alexander Amelkin <alexander@amelkin.msk.ru>
Before this commit the bytes for the GUID 'node' part were displayed in
reverse order as if they were reported by BMC according to SMBIOS
specification, not accordint to IPMI.
The field order in guid_t was also specified according to RFC4122/SMBIOS,
which is the reverse of IPMI specification. It has now been fixed.
The time_low field of GUID was taken directly, without taking in
account the ipmitool host endianness. It is now properly converted
from IPMI little-endian to host byte order. Other GUID fields are
also properly converted now. As of now, ipmitool does not generally work
properly when built for a big-endian target, but that's out of scope
of this commit.
Please note that this commit most probably breaks the output of
`ipmitool mc guid` with most existing BMC implementations, but that's
just an indication of them being broken. A follow-up commit will
re-add support for the broken behavior.
Partially resolvesipmitool/ipmitool#25
Signed-off-by: Alexander Amelkin <alexander@amelkin.msk.ru>
Add ARRAY_SIZE() macro.
Use the new macro in all places where array size is calculated by
means of sizeof(array)/sizeof(array[0]).
Fix minor bugs and warnings in the affected code.
Signed-off-by: Alexander Amelkin <alexander@amelkin.msk.ru>
