split readFRUContents() into FruUtils module

Splits the readFRUContents() method into its own module. This utility
code can then be more easily tested without extra dependencies.

Signed-off-by: Pa

split readFRUContents() into FruUtils module

Splits the readFRUContents() method into its own module. This utility
code can then be more easily tested without extra dependencies.

Signed-off-by: Patrick Venture <venture@google.com>
Change-Id: I1b8fa7179d7d0092e0236e8586f99c45121eca84

show more ...

FruDevice: Add error message for header

Verifying each bytes of header and reporting error message for
respective byte failure like version, pad or checksum under
debug for debugging.

Signed-off-by

FruDevice: Add error message for header

Verifying each bytes of header and reporting error message for
respective byte failure like version, pad or checksum under
debug for debugging.

Signed-off-by: Vijay Khemka <vijaykhemkalinux@gmail.com>
Change-Id: I9a651316b7653a46b485f10f7bbb32415ee427b4

show more ...

entity-manager: scan after ObjectManager signals

Trigger a rescan whenever interfaces are added or removed, since they
might result in a probe event.

The schema of ObjectManager signals does not al

entity-manager: scan after ObjectManager signals

Trigger a rescan whenever interfaces are added or removed, since they
might result in a probe event.

The schema of ObjectManager signals does not allow us to match the
specific interfaces in our probes so we are stuck subscribing to all
events from all services. Rescans could be avoided however, with a
custom signal handler with wrapper logic that discards wakeups without a
rescan if the interfaces added (or removed) are not in our list of probe
interfaces.

Change-Id: I834db793ebe05305248a02cb04b601ff0cc6a1ff
Signed-off-by: Brad Bishop <bradleyb@fuzziesquirrel.com>

show more ...

entity-manager: scan after NameOwnerChanged

We need a poke from DBus for static providers that create all their
objects prior to claiming a well-known name, and thus don't emit any
org.freedesktop.D

entity-manager: scan after NameOwnerChanged

We need a poke from DBus for static providers that create all their
objects prior to claiming a well-known name, and thus don't emit any
org.freedesktop.DBus.Properties signals. Similarly if a process exits
for any reason, expected or otherwise, we'll need a poke to remove
entities from DBus.

Testcase is as follows. Consider a static provider service
xyz.openbmc_project.Inventory.Manager[1] that implements interfaces found
in a DBus probe statement. Prior to this patch these commands do not
result in a successful probe:

systemctl stop xyz.openbmc_project.EntityManager
systemctl stop xyz.openbmc_project.Inventory.Manager
systemctl start xyz.openbmc_project.EntityManager
systemctl start xyz.openbmc_project.Inventory.Manager

Yet these commands do:
systemctl stop xyz.openbmc_project.EntityManager
systemctl stop xyz.openbmc_project.Inventory.Manager
systemctl start xyz.openbmc_project.Inventory.Manager
systemctl start xyz.openbmc_project.EntityManager

However if we then:
systemctl stop xyz.openbmc_project.Inventory.Manager

entity manager does not detect the missing objects and remove the
previously exposed configuration (un-probe).

With this patch applied, all scenarios above result in a successful
probe (or un-probe).

[1] To execute this testcase, this series is required:
https://gerrit.openbmc-project.xyz/38853

Change-Id: I54f5b445bafb37551664d57b0444701dc54013f7
Signed-off-by: Brad Bishop <bradleyb@fuzziesquirrel.com>

show more ...

FruDevice: improve warning messages in FRU parse logic

When got any warnings while formatFRU, is hard to understand which
device cause it since formatFRU has no information about device address.
Thi

FruDevice: improve warning messages in FRU parse logic

When got any warnings while formatFRU, is hard to understand which
device cause it since formatFRU has no information about device address.
This adds printing device address after parsing FRU data if
error/warning occurs.

Makes some check to be not critical (checksum error and non-zero bytes
after EndOfFields) since we found invalid FRU in FLEXTRONICS
S-1100ADU00-201 PSU where was wrong size in PRODUCT section header:

00000000 01 00 00 00 01 09 00 f5 01 09 19 cb 46 4c 45 58 |............FLEX|
00000010 54 52 4f 4e 49 43 53 cf 53 2d 31 31 30 30 41 44 |TRONICS.S-1100AD|
00000020 55 30 30 2d 32 30 31 ca 47 38 34 30 32 37 2d 30 |U00-201.G84027-0|
00000030 30 37 c2 30 31 cc 45 58 57 44 36 34 39 30 31 31 |07.01.EXWD649011|
00000040 32 31 c0 c0 c1 00 00 dc 00 02 18 3f a7 4c 04 28 |21.........?.L.(|
00000050 0a 37 05 28 23 b0 36 50 46 20 67 2f 3f 0a 1f f8 |.7.(#.6PF g/?...|
00000060 05 00 00 00 21 01 02 0d 6d 83 01 b0 04 74 04 ec |....!...m....t..|
00000070 04 78 00 00 00 ff ff 01 82 0d ae c2 82 b0 04 74 |.x.............t|
00000080 04 ec 04 78 00 00 00 34 08 |...x...4.|

Also fixes some other messages.

Tested: ensure there are expected messages in the log for FRU above:
fru-device[355]: Checksum error in FRU area PRODUCT
fru-device[355]: Computed checksum: 0xb0
fru-device[355]: The read checksum: 0x28
fru-device[355]: Non-zero byte after EndOfFields in FRU area PRODUCT
fru-device[355]: there were warnings while parsing FRU for device at bus 7 address 82

Change-Id: I1e1d08ca0547ec4e6a8ce4e1f71a479622d5f243
Signed-off-by: Andrei Kartashev <a.kartashev@yadro.com>

show more ...

FruDevice: print checksum values on checksum error

Following up adding verbose diagnostic on FRU decoding, this adds
printing exact checksum values against to just printing mismatch error.

Tested:

FruDevice: print checksum values on checksum error

Following up adding verbose diagnostic on FRU decoding, this adds
printing exact checksum values against to just printing mismatch error.

Tested: 1) tested with good FRU: no changes
2) tested with invalid FRU: got message in the log:
fru-device[355]: Checksum error in FRU area PRODUCT
fru-device[355]: Computed checksum: 0xb0
fru-device[355]: The read checksum: 0x28

Change-Id: Ieabc143d81a6f816351eebf0ba5ab7552a7f5ee3
Signed-off-by: Bruce Mitchell <Bruce_Mitchell@phoenix.com>
Signed-off-by: Andrei Kartashev <a.kartashev@yadro.com>

show more ...

fru-device: Fix -Werror=sign-compare

Fix a warning under GCC10. There doesn't seem to be any need for this
length holding variable to be signed.

Fixes: 2f0de17eddbc
Change-Id: If19849acd975b496992

fru-device: Fix -Werror=sign-compare

Fix a warning under GCC10. There doesn't seem to be any need for this
length holding variable to be signed.

Fixes: 2f0de17eddbc
Change-Id: If19849acd975b49699218296742c8a369e4b8af2
Signed-off-by: Brad Bishop <bradleyb@fuzziesquirrel.com>

show more ...

FruDevice: additional diagnostics when decode FRU

FruDevice currently has kind of loyal policy for FRU decoding, so one
can have a broken FRU but doesn't notice it.
This adds more diagnostics when d

FruDevice: additional diagnostics when decode FRU

FruDevice currently has kind of loyal policy for FRU decoding, so one
can have a broken FRU but doesn't notice it.
This adds more diagnostics when decode FRU binary.

Tested: 1) tested with several different good FRUs
2) tested with modified FRU like:
* wrong checksum in data area
* no C1 in the end of area
* change one of C0 to C1

Signed-off-by: Andrei Kartashev <a.kartashev@yadro.com>
Change-Id: Id703dada73eb1fa35abd531a3303b06288bd4aad

show more ...

FruDevice: try to parse Fru even it a bit broken

It was discovered that some of Delta PSUs have FRU which is not properly
formatted: there was no endOfFields mark in product area. This FRU
actually

FruDevice: try to parse Fru even it a bit broken

It was discovered that some of Delta PSUs have FRU which is not properly
formatted: there was no endOfFields mark in product area. This FRU
actually is not valid but we still can show useful information to user.

This change fixes some problems in out of area control and make decoder
not fail if endOfFields was not found but all general fields are
present.

Tested: tested with several different Frus

Signed-off-by: Andrei Kartashev <a.kartashev@yadro.com>
Change-Id: If5a386e3c2cb9412a13cc43cdf5a622c2250082d

show more ...

FruDevice: parse more than 2 custom fields

The current implementation supports up to two custom fileds for each
FRU Area but there is no such limitation in the specification.

This adds support for

FruDevice: parse more than 2 custom fields

The current implementation supports up to two custom fileds for each
FRU Area but there is no such limitation in the specification.

This adds support for unlimited number of custom fields

Tested: tested with FRU which contains three binary BOARD_INFO_AM fields

Change-Id: I6722ff6d98d27badef7a1493d67510596b6cabe9
Signed-off-by: Andrei Kartashev <a.kartashev@yadro.com>

show more ...

FruDevice: use FRU Areas enum to iterate

Using integer constants more effective and simple then iterating over
strings array.
This change refactor code that works with FRU areas to use enum.

Tested

FruDevice: use FRU Areas enum to iterate

Using integer constants more effective and simple then iterating over
strings array.
This change refactor code that works with FRU areas to use enum.

Tested: verified that there is no visible changes in behavior.

Signed-off-by: Andrei Kartashev <a.kartashev@yadro.com>
Change-Id: Ic95cddbcc3301fcf49c011a603d6373852f51058

show more ...

Add retries to mapper calls

During cycling we're seeing sometimes we exit due
to mapper errors. Put in retries to avoid EM shutting
down.

Tested: Saw retries happen and EM stay up

Change-Id: I2caa

Add retries to mapper calls

During cycling we're seeing sometimes we exit due
to mapper errors. Put in retries to avoid EM shutting
down.

Tested: Saw retries happen and EM stay up

Change-Id: I2caa4a7ca0ae17a621c23152b3c362442c45592e
Signed-off-by: James Feist <james.feist@linux.intel.com>

show more ...

FruDevice: improve updateFRUProperty function

updateFRUProperty function is used to write ASSET field but it
pretended to be generic.

This change fixes number of issues like
* wrong size check
* no

FruDevice: improve updateFRUProperty function

updateFRUProperty function is used to write ASSET field but it
pretended to be generic.

This change fixes number of issues like
* wrong size check
* no control of data overflow
* erase FRU Area next to updated area
* wrong handle of first/last fields in area
and some other
This change also implements functionality to resize FRU Area, but it is
disabled by default.

Tested: Tested by setting all properties as writable ad then trying to
change them using "set-property" busctl call to FruDevice.

Signed-off-by: Andrei Kartashev <a.kartashev@yadro.com>
Change-Id: Ic049bbf591878b4af344a352a2bec296e70cfb6e

show more ...

FruDevice: add binary fields decoding

Binary fields are generally contains binary data, not text.
This commit adds printing binary fields in HEX representation, as
follows:
BOARD_INFO_AM1: "00112233

FruDevice: add binary fields decoding

Binary fields are generally contains binary data, not text.
This commit adds printing binary fields in HEX representation, as
follows:
BOARD_INFO_AM1: "00112233"

Tested: tested with Fru contains binary custom fields

Change-Id: Iaf271e54dd02e2a00167da5657857dc49b3e9734
Signed-off-by: Andrei Kartashev <a.kartashev@yadro.com>

show more ...

FruDevice: some code cleanup

* Rename some variables to better represent meaning.
* Replace "Fru" in fuctions/variables names to "FRU" according to
OpenBMC "C++ Coding Style and Conventions"
* Rep

FruDevice: some code cleanup

* Rename some variables to better represent meaning.
* Replace "Fru" in fuctions/variables names to "FRU" according to
OpenBMC "C++ Coding Style and Conventions"
* Replace magical "8" with global named constant
* Change some variables types from int to size_t

Signed-off-by: Andrei Kartashev <a.kartashev@yadro.com>
Change-Id: If492b1b7ad3ddc097a773530767c4d9c8010b3aa

show more ...

entity-manager: remove dead code

After the probe match refactor, some code is unused.

Change-Id: Ia2e2b708db17aa7a93e8a08a3260fe3af0f42b1f
Signed-off-by: Brad Bishop <bradleyb@fuzziesquirrel.com>

entity-manager: probe: Refactor matching

In a nutshell, rely on nlohmann's comparison operator to make the match
instead of rolling our own. A nice side effect is the expected match
behavior is wel

entity-manager: probe: Refactor matching

In a nutshell, rely on nlohmann's comparison operator to make the match
instead of rolling our own. A nice side effect is the expected match
behavior is well documented by nlohmann.

Code size and performance are unaffected (after removing dead code):
before: 25121752 bytes unstripped, 3202608 bytes stripped
after: 25146552 bytes unstripped, 3202616 bytes stripped
before: ~22 seconds for 5000 iterations of the tests
after: ~22 seconds for 5000 iterations of the tests

This refactor reduces the number of lines of code we need to maintain
and test and increases the number of lines of code used from the
extensively tested nlohmann json library.

The refactor simplifies the logic down to:
return probe == value;
for non-string probes. A string specialization is still required so we
can pattern match.

The refactor enables support for matching non-trivial json types such as
arrays and objects. New types need only be added to BasicVariantType.

There are several changed testcase expectations introduced by the
refactor. Each is discussed in detail below.

- stringRegexError
- boolStringError -> boolNeqString
- falseEmptyError -> falseNeqEmpty
- trueEmptyError -> trueNeqEmpty
- uintStringError -> uintNeqString
- uintEmptyError -> unitNeqEmpty
- intStringError -> intNeqString
- intEmptyError -> intNeqEmpty
- doubleStringError -> doubleNeqString
- doubleEmptyError -> doubleNeqEmpty
Configuration file validation of regular expression syntax or the
correct json type for a given dbus interface and property is better done
by our offline validation tools. Avoid runtime complexity attempting to
distinguish between user errors and probe statements that simply do not
match by treating them the same and always return "no-match" instead of
an invalid argument error for user error situations. stringRegexError
could be an exception - there is no additional runtime complexity
overhead associated with flagging it as a user error, but doing so would
result in inconsistent behavior, so return no-match for it as well.

- stringZeroEqFalse -> stringZeroNeqFalse
- stringOneEqTrue -> stringOneNeqTrue
- stringEqUnit8 -> stringNeqUint8
- stringEqUint32 -> stringNeqUint32
- stringEqInt32 -> stringNeqInt32
- stringRegexEqInt32 -> stringRegexNeqInt32
- stringEqDouble -> stringNeqDouble
Prior to this patch dbus properties are converted to strings in the
event of a string probe. This works around mistakes in configuration
files where the wrong json type is used but it can provide unexpected
results like the probe "1" matching a true and the probe "11" matching a
false, and the string "1*2" matching the number 112. Implementing the
expected behavior is possible but would add complexity. For these
reasons this refactor drops the string conversion and requires the user
to use the correct types in their configurations.

- falseEqUint32Zero -> falseNeqUint32Zero
- trueEqUint32Zero -> falseNeqUint32Zero
- trueEqDoubleOne -> trueNeqDoubleOne
- falseEqDoubleZero -> falseNeqDoubleZero
- uintEqTrue -> uintNeqTrue
- uintEqFalse -> uintNeqFalse
- unitEqDoubleRound -> uintNeqDouble
Prior to this patch dbus properties are coerced to unsigned integers in
the event of a boolean or unsigned integer probes. Like in the previous
section this can have unexpected results such as integral json types
matching floats with fractional components from dbus. Again,
implementing the expected behavior is possible but would add
coomplexity, and again these conversions are abandoned and instead
require the user to use the correct types in their configurations.

- intEqDoubleRound -> intNeqDoubleRoound
- doubleEqFalse -> doubleNeqFalse
- doubleEqTrue -> doubleNeqTrue
Like the previous two sections, dbus properties are coerced to integers
in the event of an integer probe, and dbus properties are coerced to
doubles in the event of a double probe, and can result in unexpected
behavior. Require the user to use the correct types in their
configurations.

Change-Id: I6ce7a9c4a80d9612b15cd5e7bb4647bd96852e48
Signed-off-by: Brad Bishop <bradleyb@fuzziesquirrel.com>

show more ...

entity-manager: probe: factor out probe logic

Move the probe matching logic to a function so unit tests can be written
for it.

Change-Id: Ie6d1ec82deb48437138c9065aea3a0eeacc9cffb
Signed-off-by: Br

entity-manager: probe: factor out probe logic

Move the probe matching logic to a function so unit tests can be written
for it.

Change-Id: Ie6d1ec82deb48437138c9065aea3a0eeacc9cffb
Signed-off-by: Brad Bishop <bradleyb@fuzziesquirrel.com>

show more ...

docs: Add minimal doxygen setup

Add a Doxyfile and the minimum required tags to the sources to make
anything useful appear in the documentation.

To generate just run: doxygen from the project root

docs: Add minimal doxygen setup

Add a Doxyfile and the minimum required tags to the sources to make
anything useful appear in the documentation.

To generate just run: doxygen from the project root or doxygen
./Doxyfile.

Change-Id: Ie60eb1458586dc0a7843dca69e969b78817260d3
Signed-off-by: Brad Bishop <bradleyb@fuzziesquirrel.com>

show more ...

fru-device: explicitly include headers

We were depending on our dependencies to include asio.hpp. Our
dependency (sdbusplus) reworked things such that it doesn't include
asio.hpp and now we are bro

fru-device: explicitly include headers

We were depending on our dependencies to include asio.hpp. Our
dependency (sdbusplus) reworked things such that it doesn't include
asio.hpp and now we are broken.

Change-Id: I3feb1b30cd91534e3a4ebb605f7a25af7322aa3f
Signed-off-by: Brad Bishop <bradleyb@fuzziesquirrel.com>

show more ...

FruDevice: use uint8_t instead of char for data

Currently there is mix of char and uint8_t in binary data handling code,
which require lot of unnecessary casts.
This change replaces char by uint8_t

FruDevice: use uint8_t instead of char for data

Currently there is mix of char and uint8_t in binary data handling code,
which require lot of unnecessary casts.
This change replaces char by uint8_t everywhere and removes casts where
they are not needed anymore.

Tested: tested that there is no changes in FRU data published to dbus

Signed-off-by: Andrei Kartashev <a.kartashev@yadro.com>
Change-Id: Ic1ee984299c9be790a661e0c8e2111b3a2b98d6b

show more ...

probe: don't match unsupported json types

Prior to this patch unrecognized json types, such as arrays would cause
a probe statement to match.

Change-Id: I05575e70ad5841b988ca1ae929381011c4cec28e
Si

probe: don't match unsupported json types

Prior to this patch unrecognized json types, such as arrays would cause
a probe statement to match.

Change-Id: I05575e70ad5841b988ca1ae929381011c4cec28e
Signed-off-by: Brad Bishop <bradleyb@fuzziesquirrel.com>

show more ...

FruDevice: fix FRU 6-bit ASCII field decoding

Fix uninitialized variable in 6-bit ASCII Fru field decoder which can
cause wrong decoding of first 4 symbols group.

Signed-off-by: Andrei Kartashev <a

FruDevice: fix FRU 6-bit ASCII field decoding

Fix uninitialized variable in 6-bit ASCII Fru field decoder which can
cause wrong decoding of first 4 symbols group.

Signed-off-by: Andrei Kartashev <a.kartashev@yadro.com>
Change-Id: I82522c651344fd08e707fa6ffde0974d3b970d14

show more ...

Entity-Manager: Update product AssetTag inside Fru

Enable product assetTag property to writable in FruDevice.
Added update AssetTag method to write new assetTag.

Tested:
- Tested using "set-propert

Entity-Manager: Update product AssetTag inside Fru

Enable product assetTag property to writable in FruDevice.
Added update AssetTag method to write new assetTag.

Tested:
- Tested using "set-property" busctl call to FruDevice Service
and checked that PRODUCT_ASSET_TAG property is getting updated.

Change-Id: Iff492b6d53467246a16515c4a766fcc99f4d0f20
Signed-off-by: Joshi-Mansi <mansi.joshi@linux.intel.com>

show more ...

Fix segfault caused by accessing non existing key in busmap

Tested:
- Succesfully boot OpenBMC with error that key does not exist in
busmap.

Signed-off-by: Wludzik, Jozef <jozef.wludzik@intel.c

Fix segfault caused by accessing non existing key in busmap

Tested:
- Succesfully boot OpenBMC with error that key does not exist in
busmap.

Signed-off-by: Wludzik, Jozef <jozef.wludzik@intel.com>
Change-Id: I79736a786f9a88886ee8c54f7c6bf6e32d855769
Signed-off-by: James Feist <james.feist@linux.intel.com>

show more ...