| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
The logger changes to move to std::format incidentally caused format
errors to no longer be flagged at compile time.
The example here[1] shows that the latest gcc/c++ gave us a way to solve
this, using std::format_string.
This incidentally shows two places where we got the format arguments
wrong, so fix them.
[1] https://stackoverflow.com/questions/72795189/how-can-i-wrap-stdformat-with-my-own-template-function
Change-Id: Id884200e2c98eeaf5ef8db6c1d6362ede2ffb858
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
Clang has new checks for std::move/std::forward correctness, which
catches quite a few "wrong" things where we were making copies of
callback handlers.
Unfortunately, the lambda syntax of
callback{std::forward<Callback>(callback)}
in a capture confuses it, so change usages to
callback = std::forward<Callback>(callback)
to be consistent.
Tested: Redfish service validator passes.
Change-Id: I7a111ec00cf78ecb7d5f5b102c786c1c14d74384
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
A more in depth explanation of this is here[1], but being explicit
flushing the buffer (using std::flush) captures intent better here,
which we intend a newline, then a flush. Keeping those as separate
captures the intended behavior.
Tested: Launching bmcweb shows logging statements still function.
[1] https://clang.llvm.org/extra/clang-tidy/checks/performance/avoid-endl.html
Change-Id: I6aa427f4e4134ce30ce552cbfdd5d7be3df56c47
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
Minor syntax error injected in:
1873a04f Reduce multi-level calls of req.req members
Tested: Code compiles.
Change-Id: Iec26273349df6063eb425e4a75b1250c17bc6f3f
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
"inline static void func()"
Doesn't make sense when put in a header file. Find all instances, and
make them inline like we do everywhere else.
Tested: Code compiles.
Change-Id: I7da5821b1e372941680f82939627af39fdc2a4eb
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
Changing
static bool getUniqueEntryID
to
inline bool getUniqueEntryID
Exposes the fact that there's some undefined behavior here, and unit
tests start failing, likely due to stack being reused where it
previously wasn't.
This commit cleans up the code to simplify it, and remove the problem.
Tested: Unit tests pass. Good coverage.
Change-Id: I5b9b8e8bb83c656560193e680d246c8513ed6c02
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
The redfish-health-populate option was scheduled to be removed in 1Q
2024. It is now 2Q, so remove the option. No upstream layers enabled it
and did not find a downstream layer that did either.
This was always limited to a few resources. Overall this design was only
half done. A future "HealthRollup" can be proposed.
Some discord discussion:
[1]: https://discord.com/channels/775381525260664832/855566794994221117/1110728560819327069
Commit disabling this (merged 10 months ago):
[2]: https://github.com/openbmc/bmcweb/commit/6f8273e49cffdd347c223b9538558edfb05e818a
Tested: Code compiles
Change-Id: I4d33c1e674ecdb0fd256df62f3795073454ae7a1
Signed-off-by: Gunnar Mills <gmills@us.ibm.com>
|
|
Today, patching ethernet ip address arrays can use several
styles.
IpAddresses: [{}, {value: value}, null]
All 3 of those elements are legal. Today, we unpack values like that
with nlohmann::json, then iterate and unpack further. This leads to
problems where:
IpAddresses: [{}, {value: value}, 1.0]
would have the same behavior as the prior, given that we check for
"is_object()" to determine the null state. This is messy at best, and
not typesafe at worst.
Changing this code to use the new class NullOr<> allows the readJson
parser to fail the second example.
Change-Id: Id91f48bb64271dd568041a7c0b1ad285b59d5674
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
Use multiple level direct read.
Tested: Visual only. Need help if anyone wants to test.
Change-Id: I67c77bdd42a05a42f9cd1b40dc74517dceebdaad
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
Use multiple level direct read.
Tested: Visual only. Need help if anyone wants to test.
Change-Id: Ifcf716a2ba93fd565bbf134d4132532e60e3b4f0
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
In Redfish schema, just about all values can be a type (string,
EDM.Numeric, etc) or null. Most APIs don't allow explicitly setting
null, but there are a few cases where it is useful, namely in lists,
where an an empty object {} keeps the value the same, and null deletes
the value from the list.
Previously we handled this by unpacking as nlohmann::json, but this
allowed things like
[1.0, {}] to pass the check for an array of string values. We'd
ideally like to reject the 1.0 at the first stage, as well as reduce
the number of tiered readJson calls that we make.
This commit introducess support for unpacking std::variant types, that
allows unpacking a known type, or explicitly allowing null, by unpacking
std::nullptr_t.
Tested: Unit tests pass.
Change-Id: Ic7451877c824ac743faf1951cc2b5d9f8df8019c
Signed-off-by: Ed Tanous <edtanous@google.com>
|
|
d547d8d2c30a7d00852855da8ecc15c0cc424b0e caused a minor regression in a
poorly executed merge conflict. The reqReader instance never gets
initialized, so all payloads now fail.
Tested: Redfish service validator modified to use http2 now succeeds
when using http/2.
Change-Id: If00f5bc1d1a70396171eddd6a643fbd860e8508c
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
nlohmann::json::dump() is not an easy function to get the call
parameters correct on. We should limit the places we use it.
Luckily, both logging and redfish::messages support printing
json values directly. Use them where appropriate.
Tested: Error logging and out of range calls only of heavily used
messages and logging calls. Inspection only.
Change-Id: I57521d8791dd95250c93e8e3b2d4a959740ac713
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
Currently local meson build breaks on tinyxml2 since the code from
this package produces a warning and all warnings are treated as errors:
"""
../subprojects/tinyxml2-9.0.0/xmltest.cpp:711:38: error: ignoring
return value of ‘char* fgets(char*, int, FILE*)’ declared with
attribute ‘warn_unused_result’ [-Werror=unused-result]
"""
The warning above comes from the compilation of the tinyxml2 unit tests.
Since we don't really need them, disable 'tests' option for tinyxml2
dependency to fix the meson build.
Tested:
Locally executed "meson setup build && cd build && meson compile" now
finishes successfully.
Change-Id: I7148856a3eebbda22d03c9715955b6ab78a933fe
Signed-off-by: Konstantin Aladyshev <aladyshev22@gmail.com>
|
|
This commit changes sdbusplus setProperty calls in
log_services.hpp file to "setDbusProperty" method in Redfish
namespace that handles all DBus errors in a consistent manner.
Change-Id: Icd9b0f0326c75a1421756d515408b303bdd738e3
Signed-off-by: Asmitha Karunanithi <asmitk01@in.ibm.com>
|
|
TemporaryFileHandle class is used to create temp files in the
filesystem, and hold a handle to them until the class goes out of scope,
at which time they can be removed. It replaces makeFile(), which was
not RAII safe if an exception gets thrown, and could potentially leave
files in the filesystem if the tests fail.
Tested: Unit tests pass
Change-Id: I03eb0d342a6cd7b78115a8c42be9175f30c4ccd0
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
The redfish-enable-proccessor-memory-status option was scheduled to be
removed in 1Q 2024. It is now 2Q, so remove the option. No upstream
layers enabled it and I could not find a downstream layer that did
either.
Redfish deprecated the Processor/Memory Summary Status (state, health,
healthrollup) attributes.
Discussion on discord, when disabling:
[1]: https://discord.com/channels/775381525260664832/855566794994221117/1093939076710793296
Commit disabling this (merged 10 months ago):
[2]: https://github.com/openbmc/bmcweb/commit/5fd0aafb0f14fb3011970e8575647bb608688c7c
Tested: Code builds.
Change-Id: I539cd5f384633afa7badf1cecfc6c7a87062f672
Signed-off-by: Gunnar Mills <gmills@us.ibm.com>
|
|
Several places access the members of `req` indirectly like
`req.req.method()`. This can be simplified as `req.method()` .
This would also make the code clearer.
Tested:
- Compiles
- Redfish service validator passes
Change-Id: Ie129564ff907cdea7ac224b1e3d80cc0dedfbd7b
Signed-off-by: Myung Bae <myungbae@us.ibm.com>
|
|
Change-Id: I147664c3d181ba8ec535c7cddcb5c714e05616ea
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
Have seen defects where hitting the max connection limit with multiple
server managers attached. Although not common to exceed 100, can hit
this when using 2 or 3 webui-vue GUIs and a server manager attached.
webui-vue can use ~30 of these on its own; this isn't that hard to hit.
Nginx by default sets 512 connections[1] , so 200 for an embedded
target doesn't seem that unreasonable:
Apache sets 256 by default [2]
lighttpd sets 1024 [3]
We're in line for the defaults for other webservers.
Tested: Sent 180 basic auth requests seen bmcweb memory at
2189 2178 root R 29080 4% 49% ./bmcweb
This was on a AST2600 (p10bmc)
The connections open got to:
[DEBUG "http_connection.hpp":79] 0x19bb5c8 Connection open, total 161
Came back down as expected:
[DEBUG "http_connection.hpp":89] 0x1a41440 Connection closed, total 1
Didn't see this with multiple webui-vues / server managers.
[1] https://nginx.org/en/docs/ngx_core_module.html#worker_connections
[2] https://httpd.apache.org/docs/2.4/mod/mpm_common.html#maxrequestworkers
[3] https://redmine.lighttpd.net/projects/1/wiki/Server_max-connectionsDetails
Change-Id: I807302e32e61e31212850a480d721d89d484593f
Signed-off-by: Gunnar Mills <gmills@us.ibm.com>
|
|
This is the same version Yocto uses.
[1] https://github.com/openbmc/openbmc/blob/8c1713b6664523ff7a5bc300ca00c9cde8e5b2c1/poky/meta/recipes-support/nghttp2/nghttp2_1.60.0.bb
Change-Id: I71485c559f37dc21cabeab6b95e38c8a30073af5
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
These tests are commented out, and have been for a very long time.
Clearly they don't matter.
Change-Id: I084378ee9bc43bb64bd6e134398bbf2173d263ff
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
Clang-tidy-18 has new checks that can find more cases where we've
missed an opportunity to std::move.
Fix them.
Tested: Logging works, unit tests pass.
Change-Id: I0cf58204ce7265828693b787a7b3a16484c3d5e5
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
readability-avoid-nested-conditional-operator
With one exception, we already pass this check. Update the log services
code to make it pass, and update it to use the generated enums.
Tested: Code inspection only.
Change-Id: Ic80a7382beb0f541de4916d7b51e42ed5d5dc542
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
This commit changes sdbusplus setProperty calls in
hypervisor_system.hpp file to "setDbusProperty" method in Redfish
namespace that handles all DBus errors in a consistent manner.
Change-Id: Iebca5eb4e28159d61cd4b13c0343b78efd0f1f39
Signed-off-by: Asmitha Karunanithi <asmitk01@in.ibm.com>
|
|
This include exists above.
Tested: code compiles.
Change-Id: I8e5d7bce292486d2f534da1b539212113c1e8d56
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
When running compiling from local dependencies, it's nice to be able to
move binaries around wherever they're needed.
When we moved to the cmake version of boost, we started building shared
libraries on accident. nghttp2 has always pulled shared libraries since
introduction.
Go back to building static libraries.
Tested: Binary build from desktop launches directly without a
"missing boost_url.so" error.
Change-Id: I9858f27f21841097bd03a6bbcac953b08caa4b14
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
inline is not required on member methods. Clang-tidy has a check for
this. Enable the check and fix the two bad usages.
Tested: Code compiles.
Change-Id: I3115b7c0c4005e1082e0005b818fbe6569511f49
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
clang-18 improves this check so that we can actually use it. Enable it
and fix all violations.
Change-Id: Ibe4ce19c423d447a4cbe593d1abba948362426af
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
Redfish protocol validatator has SSE tests that expose some bad coding
practies in SSE handlers, namely, that there are several cases where we
don't check for nullptr.
Fix them.
This appears to have been introduced in:
https://gerrit.openbmc.org/c/openbmc/bmcweb/+/41319
Tested: Redfish service validator passes more tests.
Change-Id: Id980725f007d044b7d120dbe0f4b625865cab6ba
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
The commit:
b5f288d Make use of filebody for dump offload
Caused a minor failure in clearing responses, where open file handles
wouldn't be closed in between queries, resulting in the next read to
return empty content. This caused redfish protocol validator to fail.
boost::beast::http::response::clear() documentation shows that it only
clears the headers, not the file body. Now normally, this doesn't
matter, because bmcweb completely replaces the response body when a new
response is driven, but not in the case of files.
Add response.body().clear() during the clear to ensure the response is
cleared.
In addition, add encodingType to the clear() call, to ensure that it is
reset as well. This is a bug, but I don't know the reproduction steps.
Tested: Redfish protocol validator now completes (with SSE failures)
Change-Id: Ice6d5085003034a1bed48397ddc6316e9cd0536f
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
When we have an internal error, having the D-Bus response is really
helpful. Follow our guide and have these be a Logging Level Error.
Tested: None. Inspection only.
Change-Id: Ie1d9f364c3af7f2a8839d878d68c82c10ddc0429
Signed-off-by: Gunnar Mills <gmills@us.ibm.com>
|
|
Most of this code was written before bmcweb had a logger, and
therefore used cout/cerr.
This commit greps the codebase and finds all places where we still
use cout/cerr, and moves them to logging.
Tested: Inspection only. No functional changes.
Change-Id: I5ce1883c9941e80203ec29decb3a0206fd118506
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
This commit changes sdbusplus setProperty calls in ethernet.hpp file
to "setDbusProperty" method in Redfish namespace that handles all DBus
errors in a consistent manner.
Tested By:
Tested a few PATCH operations on the redfish endpoints defined in
this file and verified that bmcweb returns appropriate Redfish
errors.
Change-Id: Ie456db75d59dc247cdce5dd5cc0b2f6894f5265f
Signed-off-by: Asmitha Karunanithi <asmitk01@in.ibm.com>
|
|
There are currently 78 sdbusplus::asio::setProperty calls in
redfish-core. The error handler for nearly all of them looks something
like:
```
if (ec)
{
const sd_bus_error* dbusError = msg.get_error();
if ((dbusError != nullptr) &&
(dbusError->name ==
std::string_view(
"xyz.openbmc_project.Common.Error.InvalidArgument")))
{
BMCWEB_LOG_WARNING("DBUS response error: {}", ec);
messages::propertyValueIncorrect(asyncResp->res, "<PropertyName>", <PropertyValue>);
return;
}
messages::internalError(asyncResp->res);
return;
}
messages::success(asyncResp->res);
```
In some cases there are more errors handled that translate to more error
messages, but the vast majority only handle InvalidArgument. Many of
these, like the ones in account_service.hpp, do the error handling in a
lambda, which causes readability problems. This commit starts to make
things more consistent, and easier for trivial property sets.
This commit invents a setDbusProperty method in the redfish namespace
that tries to handle all DBus errors in a consistent manner. Looking
for input on whether this will work before changing over the other 73
calls. Overall this is less code, fewer inline lambdas, and defaults
that should work for MOST use cases of calling an OpenBMC daemon, and
fall back to more generic errors when calling a "normal" dbus daemon.
As part of this, I've ported over several examples. Some things that
might be up in the air:
1. Do we always return 204 no_content on property sets? Today there's a
mix of 200, with a Base::Success message, and 204, with an empty body.
2. Do all DBus response codes map to the same error? A majority are
covered by xyz.openbmc_project.Common.Error.InvalidArgument, but there
are likely differences. If we allow any daemon to return any return
code, does that cause compatibility problems later?
Tested:
```
curl -k --user "root:0penBmc" -H "Content-Type: application/json" -X PATCH -d '{"HostName":"openbmc@#"}' https://192.168.7.2/redfish/v1/Managers/bmc/EthernetInterfaces/eth0
```
Returns the appropriate error in the response
Base.1.16.0.PropertyValueIncorrect
Change-Id: If033a1112ba516792c9386c997d090c8f9094f3a
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
And fix the includes that are wrong.
Note, there is a very large ignore list included in the .clang-tidy
configcfile. These are things that clang-tidy doesn't yet handle
well, like knowing about a details include.
Change-Id: Ie3744f2c8cba68a8700b406449d6c2018a736952
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
Current code has
//clang-format on
When it should have
// clang-format on
The difference is subtle, but disables formatting for this whole file.
Re-enable and fix the couple of problems.
Tested: Whitespace only.
Change-Id: Ia155226327d4d611eb2c0f5232274459866e81cc
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
These classes accidentally overlapped in naming with the nghttp2
classes. This is because this class, unlike most nghttp2 classes
doesn't end in _ptr for a type. This changes the class name to add a
_ex to differentiate the two classes, and avoid a warning in clang.
Tested: Unit tests pass. Code only used in unit test.
Change-Id: I91a6982264df69bc65166ab38feddc21f72cd223
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
Enable the checks we pass already.
This also removes the commented out misc-no-recursion, considering we
don't pass it.
Tested: Clang-tidy passes.
Change-Id: Ibaed95677aed85188bff483d2cd53605faaf7cc6
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
Another clang version, another set of checks we can enable.
bmcweb passes all these checks today, so enable them to help
folks write better code.
Change-Id: Ied6a364ee92d8d634edea717cfa2fb5245d534f9
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
clang-tidy-18 makes this feature stable enough for us to use in general.
Enable the check, and fix the couple of regressions that have snuck in
since we last ran the check.
Tidy seems to not be able to understand that ASSERT will not continue,
so if we ASSERT a std::optional, it's not a bug. Add explicit checks to
keep tidy happy.
Tested: clang-tidy passes.
Change-Id: I0986453851da5471056a7b47b8ad57a9801df259
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
It's not clear how this came to be the way it is, but tidy now warns
that this variable is unused (which it is).
Refactor the LDAP code to not use the variable, and to use concrete
object_t and array_t
Tested: Redfish service validator passes.
Change-Id: I0c106d10594a396d506bf9865cb29d4a10a372a1
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
Extend the hasIndicatorLed array and add
xyz.openbmc_project.Inventory.Item.Chassis interface.
Tested:
```
curl -k https://$bmc/redfish/v1/Chassis/chassis
{
"@odata.id": "/redfish/v1/Chassis/chassis",
"@odata.type": "#Chassis.v1_22_0.Chassis",
"Actions": {
"#Chassis.Reset": {
"@Redfish.ActionInfo": "/redfish/v1/Chassis/chassis/ResetActionInfo",
"target": "/redfish/v1/Chassis/chassis/Actions/Chassis.Reset"
}
},
"ChassisType": "RackMount",
"Id": "chassis",
"Links": {
"ComputerSystems": [
{
"@odata.id": "/redfish/v1/Systems/system"
}
],
"ManagedBy": [
{
"@odata.id": "/redfish/v1/Managers/bmc"
}
]
},
"Location": {
"PartLocation": {
"ServiceLabel": "U78DA.ND0.WZS004K"
}
},
"IndicatorLED": "Off",
"LocationIndicatorActive": false,
"Manufacturer": "",
"Model": "23",
"Name": "chassis",
"PCIeDevices": {
"@odata.id": "/redfish/v1/Systems/system/PCIeDevices"
},
"PartNumber": "",
"Power": {
"@odata.id": "/redfish/v1/Chassis/chassis/Power"
},
"PowerState": "Off",
"Sensors": {
"@odata.id": "/redfish/v1/Chassis/chassis/Sensors"
},
"SerialNumber": "",
"Status": {
"Health": "OK",
"HealthRollup": "OK",
"State": "StandbyOffline"
},
"Thermal": {
"@odata.id": "/redfish/v1/Chassis/chassis/Thermal"
}
}
```
Signed-off-by: George Liu <liuxiwei@ieisystem.com>
Change-Id: I02e77d56e555f9aee3f76015baeebbf1f4a292ab
|
|
These variables aren't used, and clang-tidy-18 flags it. Remove
Tested: Code compiles.
Change-Id: I414c4614a5f789aecab7700a4ec805e98c09cade
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
bugprone-multi-level-implicit-pointer-conversion is something that we
pass currently, with one exception in the deprecated rest API. Ignore
the one exception, as it's not clear how to fix it, and enable the
check.
Tested: Clang tidy passes.
Change-Id: Idc10e0bb7b876e1c70afa28f9c27cc7bef1db0d7
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
clang-tidy-18 must've fixed their checking for these in headers.
Resolve as the robot commands.
Tested: Noop changes made by tidy. Code compiles.
Change-Id: I1de7686c597deffb0df91c30dae1a29f9ba7900e
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
Change-Id: I0133bbd0a7573bd3d1e3c3c99382442b286696f6
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
Quite often do I compile this project, and see an error message that
makes no sense. Multiple times I've seen posted about compiler errors
that amount to using an old version of clang or gcc.
Explicitly require clang-17 and gcc-13 in the meson config, and give
better errors if they're not present.
This also allows simplifying our warning flags (which probably need
a review soon) by making two sets of flags, one for each compiler.
Note, clang has the -Weverything flag, which we use, so explicitly
enabling warnings isn't really required, only disabling the ones
that we don't use.
Tested: Code compiles.
Change-Id: I09fa74e6d36feaf05710a4bb7d266f80ff1cc592
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
This code was copied from one of the nghttp2 examples, and makes things
more complicated than they should be. We only handle one case here, so
a pattern of returning early is easier.
Also, this resolves a possible clang-tidy bugprone warning (that we
don't yet enable).
Tested: Http2 unit tests pass (good coverage for this case).
Change-Id: Ie8606872f3a96f1bb0329bf22a4f7429f431bbef
Signed-off-by: Ed Tanous <ed@tanous.net>
|
|
If one sets the XSS policy disable, and tries to load the webui, they're
met with the following error message:
```
chunk-vendors.6cfb4b74.js:36 Refused to load the image
'data:image/svg+xml;charset=utf-8,%3Csvg
xmlns='http://www.w3.org/2000/svg' width='4' height='5'%3E%3Cpath
fill='%233f3f3f' d='M2 0L0 2h4zm0 5L0 3h4z'/%3E%3C/svg%3E' because it
violates the following Content Security Policy directive: "img-src *".
Note that '*' matches only URLs with network schemes ('http', 'https',
'ws', 'wss'), or URLs whose scheme matches `self`'s scheme. The scheme
'data:' must be added explicitly.
```
Do as it asks, and add data: to the content security policy.
Tested: Browser console no longer shows error when XSS is enabled.
Change-Id: I17f70d7c87a284b33ef6eb5a01a01c23a14898c9
Signed-off-by: Ed Tanous <ed@tanous.net>
|
