| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
Adds aggregation support for resource collections that take the form
of "/redfish/v1/<resource collection>". Collection URIs are
identified by the precense of a "Members" array in the response.
Resources from satellite BMCs are added to the "Members" array of
the response and the "Members@odata.count" value is updated to
denote the new array size.
These satellite resource URIs that are added also include the
prefix associated with that satellite.
Note that as a first step this patch assumes a single satellite BMC.
There are some potential race conditions that could occur for setups
with multiple satellite BMCs. This has been commented in the code
and is better left to its own patch.
Tested:
Queried various collection URIs and the aggregated resources
appeared in the response's "Members" array.
Querying 'localhost:80/redfish/v1/Chassis?$expand=.($levels=1)'
resulted in $expand correctly returning the outputs from querying
the URIs of all local and satellite Chassis resources. This would
have failed if the satellite Chassis resources were omitted from the
"Members" array or the satellite's prefix was not correctly added to
the URI.
Also queried a collection URI that only existed on the satellite BMC.
The AsyncResp was completely overwritten by the response from the
satellite BMC.
Queries to non-collection URIs resulted in no attempts to add
satellite responses to the AsyncResp.
Signed-off-by: Carson Labrado <clabrado@google.com>
Signed-off-by: Ed Tanous <edtanous@google.com>
Change-Id: I3b379cd57e5a121eb4a344d88fc8e43170ca78a6
|
|
URIs in the responses returned with Redfish Aggregation enabled will
potentially be incorrect since ones from satellite BMCs will not
include the associated prefix such as "5B247A_" in the resource ID
portion of the URIs.
This patch fixes those links so that they include their BMC's
associated prefix. Note that a future patch will be needed to add
prefixes to aggregated resources that would appear under collection
URIs such as "/redfish/v1/Chassis".
Tested:
Requests were sent to URIs associated with the aggregating BMC and a
satellite BMC denoted as "5B247A". The URIs in the responses
were successfully updated such that "5B247A_" was added for
satellite resources.
Signed-off-by: Carson Labrado <clabrado@google.com>
Signed-off-by: Ed Tanous <edtanous@google.com>
Change-Id: Ib4f976fab1ca1e8603f7cf55292732ffb71cd03e
|
|
Adds ability to route requests to either native resources or
resources that belong to a satellite BMC as part of Redfish
Aggregation. A prefix in the URI denotes if the resource is actually
from a satellite BMC. Prefixes are only used to denote satellite
resources. The URI of resources on the local/aggregating BMC will
remain unchanged.
Prefixes are separated from the resource ID by an underscore. This
means that underscores cannot be used in the prefix name itself.
The prefixes used by satellite BMCs are revealed via D-Bus as well as
the config information needed to connect to that BMC.
Requests for satellite resources will not be handled locally.
Care should be taken to not name any local resources in a way that
could cause a collision (e.g. having a Chassis object named
"aggregated0_1U" on the aggregating BMC).
The patch only covers routing requests. Requests to collection URIs
like /redfish/v1/Chassis will only return resources on the local BMC.
A future patch will cover adding satellite resources to collections.
Also note that URIs returned in the responses will not have the proper
prefix included. Fixing these URIs will be addressed in future
patches.
A number of TODO comments are included in the code to indicate that
this functionality (collections and URI fixup) still needs to be
implemented.
Example URIs w/o Redfish Aggregation:
/redfish/v1/Chassis/1U/
/redfish/v1/Systems/system/
/redfish/v1/Managers/bmc/
Example URIs after enabling Redfish Aggregation if the associated
resources are located on the local/aggregating BMC:
/redfish/v1/Chassis/1U/
/redfish/v1/Systems/system/
/redfish/v1/Managers/bmc/
Example URIs if resources are instead located on a satellite BMC
named "aggregated0":
/redfish/v1/Chassis/aggregated0_1U/
/redfish/v1/Systems/aggregated0_system/
/redfish/v1/Managers/aggregated0_bmc/
Tested:
I was able to query supported resources located on the local BMC
as well as on a satellite BMC. Requests with unknown prefixes return
a 404. Requests to resource collections only return the resources
that are located on the aggregating BMC.
Signed-off-by: Carson Labrado <clabrado@google.com>
Signed-off-by: Ed Tanous <edtanous@google.com>
Change-Id: I87a3deb730bda95e72ecd3144ea40b0e5ee7d491
|
|
Sometimes it is desirable to only parse a portion of a URL, and allow
any elements at the end. This comes up significantly in aggregation
where parsing "/redfish/v1/<collection>/anything is pretty common.
This commit adds a new class, OrMorePaths, that can be used as a
placeholder to indicate that more paths should be accepted.
Tested: Unit tests pass.
Signed-off-by: Ed Tanous <edtanous@google.com>
Change-Id: If4fb3991a91560fd3b8b838f912aa36e79ddd2b3
|
|
A continuation of 4d94272fe54c147974f86788092bbbdd950406aa,
increases the size of httpReadBodyLimit to 2^17 bytes (about 128 KB).
This is because with Redfish Aggregation we need to be able to handle
larger response sizes such as json schema files. The ComputerSystem
schema in particular is over 120 KB.
Going forward we will not be able to keep increasing the limit.
We need a better solution for handling larger response that are
larger than httpReadBodyLimit.
Tested:
Used https://gerrit.openbmc.org/c/openbmc/bmcweb/+/53310/68 to
successfully retrieve
/redfish/v1/JsonSchemas/ComputerSystem/ComputerSystem.json from a
satellite BMC.
Signed-off-by: Carson Labrado <clabrado@google.com>
Change-Id: Ida80b8fddbd1df1310d18a77ecfb44b7fdf292ef
|
|
We've accumulated several time utility functions in the http classes.
Time isn't a core HTTP primitive, so http is not where those functions
below.
This commit moves all the time functions from the crow::utility
namespace into the redfish::time_utils namespace, as well as moves the
unit tests.
No code changes where made to the individual functions, with the
exception of changing the namespace on the unit tests.
Tested: Unit tests pass.
Signed-off-by: Ed Tanous <edtanous@google.com>
Change-Id: I8493375f60aea31899c84ae703e0f71a17dbdb73
|
|
Supporting higher precision on our timestamps seems worthwhile, and
useful to logging implementations for getting more accurate numbers from
redfish LogService interfaces.
This commit updates the code to add millisecond and microsecond
precision to timestamps.
Tested: Unit tests pass, pretty good coverage here.
Signed-off-by: Ed Tanous <edtanous@google.com>
Change-Id: I0914908966702a6cf1bcb59f22761dc24c46b4c3
|
|
Found these when I went through the code path of authx.
Tested: comment only changes.
Signed-off-by: Nan Zhou <nanzhoumails@gmail.com>
Change-Id: Id80725f4bf5f3972e975347dcac8598e2ffab332
|
|
There are two overloads of addHeader, one that takes a string, and one
that takes a boost enum. For most common headers, boost contains a
string table with all of those entries anyway, so there's no point in
duplicating the strings, and ensures that we don't make trivial
mistakes, like capitalization or - versus underscore that aren't caught
at compile time.
Tested:
This saves a trivial amount (572 bytes) of compressed binary size.
curl --insecure -vvv --user root:0penBmc https://192.168.7.2/redfish/v1
returns < Content-Type: application/json
curl --insecure -vvv -H "Accept: text/html" --user root:0penBmc https://192.168.7.2/redfish/v1
Returns
< Content-Type: text/html;charset=UTF-8
Signed-off-by: Ed Tanous <edtanous@google.com>
Change-Id: I34c198b4f9e219247fcfe719f9b3616d35aea3dc
|
|
Per the coding standard, if we can support what we need to do with
std variants of something, we should prefer that. This commit adds an
iso8160 to string method that supports any arbitrary
std::chrono::duration object, which allows doing the full range of all
of our integer types, and reduces the complexity (and presumably compile
times) not pulling in a complex library.
Despite the heavy templating, this only appears to add 108 bytes of
compressed binary size to bmcweb. This is likely due to the decreased
complexity compared to the boost variant (that likely pulls in
boost::locale). (Ie 3 template instantiations of the simple one take
about the same binary space as 1 complex instantiation).
Tested:
Unit tests pass (pretty good coverage here)
Signed-off-by: Ed Tanous <edtanous@google.com>
Change-Id: I78200fb391b601eba8d2bfd2de0dd868e4390d6b
|
|
There is a bug where, when running an expand query, headers from the
response object get dropped. These headers include OData.type, and the
newly minted Link header, as well as possible others.
This was actually noted in a TODO, although the author of the TODO,
didn't fully understand the consequences at the time, and thought there
was no functional impact.
To resolve this, this commit resolves the TODO, and allows the Response
object to be moved out, instead of having to create a new one, which
preserves all the response state. To do this, it creates a move
constructor on the Response object for this use. The move constructor
is relatively benign, with one caveat, that we might be moving while in
a completion handler (as is the most common use). So both the existing
operator= and Response() move constructor are amended to handle this
case, and simply null out the response object in the copied object,
which would be correct behavior, given that each callback handler should
only be called once per Response object.
Tested:
curl --insecure --user root:0penBmc -vvvv
https://192.168.7.2/redfish/v1\?\$expand\=\*\(\$levels\=2\)
returns the same body as previously, now with the included:
OData-Version: 4.0
Allow: Get
headers in the response.
Signed-off-by: Ed Tanous <edtanous@google.com>
Change-Id: I221364dd4304903b37cacb1386f621b073a0a891
|
|
The jsonMode method is a single line, that basically only sets the
content type, and is only called from one place. In all other cases we
set the content-type directly from a header, so we should do the same
here.
Tested:
curl --insecure -vvv --user root:0penBmc https://192.168.7.2/redfish/v1
returns < Content-Type: application/json
Signed-off-by: Ed Tanous <edtanous@google.com>
Change-Id: I82f6fcf1b574192620ac7b29f97cdd01334c18c4
|
|
clang14 doesn't compile because of
"moving a temporary object prevents copy elision".
This also alligns the plaintext socket with style of SSL socket.
Tested: trivial change. It builds.
Signed-off-by: Nan Zhou <nanzhoumails@gmail.com>
Change-Id: I9203cf162d738290306f9ba73ec0ab8f2ca5033c
|
|
The Redfish standard in section 8.2 states:
A Link header containing rel=describedby shall be returned on GET and
HEAD requests for Redfish resources. If the referenced JSON Schema is a
versioned schema, it shall match the version contained in the value of
the @odata.type property returned in this resource.
This commit attempts to add this capability to ServiceRoot. Future
similar patches will start adding this across the tree.
To do this, a few things happen. First, this removes the implicit HEAD
handling in the router. Because we now need explicit HEAD handling
per-route with specific headers, there's no good way to make this
generic.
Next, it rearranges the code such that handleServiceRootGet can first
call handleServiceRootHead, to avoid duplicating the addHeader call.
Tested: Redfish protocol validator passes the
RESP_HEADERS_REL_LINK_DESCRIBED_BY check for ServiceRoot.
Signed-off-by: Ed Tanous <edtanous@google.com>
Change-Id: I92248089a3545432c14f551309ea62332e554647
|
|
These changes are done by running iwyu manually under clang14.
Suppressed some obvious impl or details headers. Kept the recommended
public headers.
IWYU can increase readability, make maintenance easier, and avoid errors
in some cases. See details in
https://github.com/include-what-you-use/include-what-you-use/blob/master/docs/WhyIWYU.md.
This commit also uses its best effort to correct obvious errors through
iwyu pragma.
See reference here:
https://github.com/include-what-you-use/include-what-you-use#how-to-correct-iwyu-mistakes
Tested: unit test passed.
Signed-off-by: Nan Zhou <nanzhoumails@gmail.com>
Change-Id: I983b6f75601707cbb0f2f04546c3362ff4ba7fee
|
|
Per the coding standard, now that C++ supports std::string::starts_with
and std::string::ends_with, we should be using them over the boost
alternatives. This commit goes through and updates all usages.
Arguably some of these are incorrect, and instances of common error 13,
but because this is mostly a mechanical it intentionally doesn't try to
handle it.
Tested: Unit tests pass.
Signed-off-by: Ed Tanous <edtanous@google.com>
Change-Id: Ic4c6e5d0da90f7442693199dc691a47d2240fa4f
|
|
If a destination is not reachable, then the connection will timeout
in doConnect(). This causes issues later when the client attempts to
resend the message.
The error check in doClose() should not exit early since that will
result in the connection's status not being marked as closed and
thus it will never get reused.
Similarly, doCloseAndRetry() should not exit early since that will
cause the retry flow to hang and the connection's callback function
will not get deleted.
Tested:
Used Redfish Aggregation patches in the chain through
https://gerrit.openbmc.org/c/openbmc/bmcweb/+/54896 to verify that
requests to collections such as /redfish/v1/Chassis no longer hang
when the specified Satellite BMC does not exist
Signed-off-by: Carson Labrado <clabrado@google.com>
Change-Id: Ic8369b0de8efb00ff168bc1ed43f1d7fd6c7366a
|
|
cppcheck takes a little issue with this logic having some redundancies
in it. Regardless of that, it's kind of hard to read; Rearrange the
logic so it's easier to read and add comments.
Tested: Redfish service validator passes.
Signed-off-by: Ed Tanous <edtanous@google.com>
Change-Id: I0251ceb511e1bc62260b68c430b272d02b90fcb7
|
|
cppcheck correctly notes that a lot of variables in the new code can be
const. Make most of them const.
Tested: WIP
Signed-off-by: Ed Tanous <edtanous@google.com>
Change-Id: I8f37b6353fd707923f533e1d61c5b5419282bf23
|
|
A number of methods in http::Response were not marked const when they
should've been. This is generally not an issue, as most usages of
Response are in a non-const context, but as we start using const
Response objects more, we need to be more careful about const
correctness.
Tested: Unit tests pass.
Signed-off-by: Ed Tanous <edtanous@google.com>
Change-Id: I8b31e71b6594d9328f106e1367084db42b783b6c
|
|
There's lots of magic numbers in this file that we inherited from crow.
This commit Adds a TypeCode enum class that can be used in place of the
magic numbers. It keeps the same values as were present previously
(0-6) in case there are places where this abstraction leaked out, but I
believe this catches all of them.
Tested: Redfish service validator passes.
Signed-off-by: Ed Tanous <edtanous@google.com>
Change-Id: I063955adb8bf75d9bb6298e29e6e44c210ee9cc3
|
|
HttpClient these days is intended to be a generic feature. It should
not be relying directly on Redfish messages. In this case, we only ever
rely on the return code internally, so replace messages with a simple
bad_gateway.
Tested: Code compiles. Unit tests pass.
Signed-off-by: Ed Tanous <edtanous@google.com>
Change-Id: Ie4bd65dc0b90b75f61ab0e31ca535eefbf0f4ebb
|
|
cppcheck correctly notes this parameter can be const.
Unfortunately we've making a copy here, but looking at the ownership,
that can't be avoided simply without a shared_ptr, and that's far more
invasive.
Signed-off-by: Ed Tanous <edtanous@google.com>
Change-Id: Ib70c89a7a1bc7210219052ba2a44bc7a1c20c7f2
|
|
This variable was clearly for the print statement below, but then got
abandoned at some point. Clean it up.
cppcheck found this as well.
Tested: Unused code
Signed-off-by: Ed Tanous <edtanous@google.com>
Change-Id: I2d143f00dc6956c87ba77d5cbe1b96be631ca795
|
|
cppcheck correctly finds that these branches will always be hit, so the
branch is unneeded. Let's start by getting the code cleaned up.
Tested: CI only, cpp check.
Signed-off-by: Ed Tanous <edtanous@google.com>
Change-Id: I7b89337a81e676915243d5bc9d26c24a89c74aef
|
|
Include what you use.
Tested: Code compiles
Signed-off-by: Ed Tanous <edtanous@google.com>
Change-Id: I23c89fee6f3e39d2f2a7dc1c3ba2db8819e8adc7
|
|
We essentially follow this rule already, not relying on implicit
operators, although there are a number of cases where in theory we
could've implicitly constructed an object.
This commit enables the clang-tidy check.
Tested: Code compiles, passes clang-tidy.
Signed-off-by: Ed Tanous <edtanous@google.com>
Change-Id: Ia428463313b075c69614fdb326e8c5c094e7adde
|
|
This patchset is the conclusion of a multi-year effort to try to fix
shadowed variable names. Variables seem to be shadowed all over, and in
most places they exist, there's a "code smell" of things that aren't
doing what the author intended.
This commit attempts to clean up these in several ways by:
1. Renaming variables where appropriate.
2. Preferring to refer to member variables directly when operating
within a class
3. Rearranging code so that pass through variables are handled in the
calling scope, rather than passing them through.
These patterns are applied throughout the codebase, to the point where
-Wshadow can be enabled in meson.build.
Tested: Code compiles, unit tests pass. Still need to run redfish
service validator.
Signed-off-by: Ed Tanous <edtanous@google.com>
Change-Id: If703398c2282f9e096ca2694fd94515de36a098b
|
|
Add using directives and namespaces.
Teseted: unit test passed.
Signed-off-by: Nan Zhou <nanzhoumails@gmail.com>
Change-Id: I4c04f07484a2c1ce29ae26b523d62071ea03996c
|
|
IWYU. Use <> for dependency headers and "" for bmcweb headers.
Reference:
https://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines#sf12-prefer-the-quoted-form-of-include-for-files-relative-to-the-including-file-and-the-angle-bracket-form-everywhere-else
Signed-off-by: Nan Zhou <nanzhoumails@gmail.com>
Change-Id: I02537f16974ab96a1e4e4e1ee6e1db1a9c679c68
|
|
Removed unnecessary using directives. Add necessary using directives.
Tested: unit test passed.
Signed-off-by: Nan Zhou <nanzhoumails@gmail.com>
Change-Id: I36948b531c4e75dbae6837601894839b23b2d3d1
|
|
IWYU. Use <> for dependency headers and "" for bmcweb headers.
Reference:
https://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines#sf12-prefer-the-quoted-form-of-include-for-files-relative-to-the-including-file-and-the-angle-bracket-form-everywhere-else
Signed-off-by: Nan Zhou <nanzhoumails@gmail.com>
Change-Id: I0338d475990eaf2f0f8105b01d63c2d9bd9fc446
|
|
Testing of https://gerrit.openbmc.org/c/openbmc/bmcweb/+/53310 revealed
that recvMessage() was failing because the body limit was being
exceeded. httpReadBodyLimit was being used to set both the body limit
as well as the buffer size. Add a different variable to set the buffer
size so that the body limit can be doubled to 16K while the fixed
buffer can be set as a smaller size of 4K.
Tested:
Queries using the mentioned patch stopped failing after applying this
change.
Signed-off-by: Carson Labrado <clabrado@google.com>
Change-Id: I33b62dfc12ff078fba9fb92b2f95a05f41acce3d
|
|
Allows individual retry policies to specify what HTTP response codes
are considered valid. Sets functions for the EventService and
Redfish Aggregation retry policies. Those functions expect a
response code and return an error code based on what the response
code is.
This change is needed because EventService only considers 2XX codes
to be valid. Any code outside of that range would trigger a retry
attempt. Redfish Aggregation by design will need to return
errors outside of that range such as 404. It should not retry to
send a message when it receives a 404 from a satellite BMC.
Right now 404 is the only error code that is handled differently
between the services. Going forward, Redfish Aggregation will
likely want to allow other error codes as its functionality is
expanded.
Tested:
Used Redfish-Event-Listener with ssh port forwarding to create 3
subscriptions. I then closed the ssh connection and sent a test
event. Bmcweb made 3 retry attempts for each subscription. At
that point the max retry amount (as defined by EventService) was
reached and bmcweb stop attempting to resend the messages.
There were no errors when the Redfish-Event-Listener was correctly
connected. Test events resulted in messages being sent for each
subscription.
Signed-off-by: Carson Labrado <clabrado@google.com>
Change-Id: Ifdfaf638d28982ed18998f3ca05280a288e0020a
|
|
While working on tests, I found that |app.hpp| is missing some boost
headers. I added them manually in this commit.
Tested: code compiles.
Signed-off-by: Nan Zhou <nanzhoumails@gmail.com>
Change-Id: I1d2fb0f312e1810d836c986e320263a9581f13f2
|
|
|useSsl| was not referenced anywhere else, so delete it.
Tested: code compiles.
Signed-off-by: Nan Zhou <nanzhoumails@gmail.com>
Change-Id: Ie6ba9c122ab0897254ed3cb0e278d43e8b6a283c
|
|
bmcweb is in a weird position where, on the one hand, we would like to
support Redfish to the specification, while also supporting a secure
webui. For better or worse, the webui can't currently use non-cookie
auth because of the impacts to things outside of Redfish like
websockets.
This has lead to some odd code in bmcweb that tries to "detect" whether
the browser is present, so we don't accidentally pop up the basic auth
window if a user happens to get logged out on an xhr request. Basic
auth in a browser actually causes CSRF vulnerabilities, as the browser
caches the credentials, so we don't want to make that auth method
available at all.
Previously, this detection was based on the presence of the user-agent
header, but in the years since this code was originally written, a
majority of implementations have moved to sending a user-agent by
default, which makes this check pretty much useless for its purpose. To
work around that, this patchset relies on the X-Requested-With header,
to determine if a json payload request was done by xhr. In theory, all
browsers will set this header when doing xhr requests, so this should
provide a "more correct" solution to this issue.
Background:
https://en.wikipedia.org/wiki/List_of_HTTP_header_fields
"X-Requested-With Mainly used to identify Ajax requests (most JavaScript
frameworks send this field with value of XMLHttpRequest)"
Tested:
curl -vvvv --insecure https://192.168.7.2/redfish/v1/SessionService/Sessions
Now returns a WWW-Authenticate header
Redfish-protocol-validator now passes 7 more tests from the
RESP_HEADERS_WWW_AUTHENTICATE category.
Launched webui-vue and logged in. Responses in network tab appear to
work, and data populates the page as expected.
Used curl to delete redfish session from store with
DELETE /redfish/v1/SessionService/Sessions/<SessionId>
Then clicked an element on the webui, page forwarded to login page as
expected.
Opened https://localhost:8000/redfish/v1/CertificateService in a
browser, and observed that page forwarded to the login page as it
should.
Signed-off-by: Ed Tanous <edtanous@google.com>
Change-Id: I60345caa41e520c23fe57792bf2e8c16ef144a7a
|
|
The usual updates to make code compile on clang again. Extra semicolons
that have snuck in, missing inline and static definitions.
Tested: Code compiles on clang.
Signed-off-by: Ed Tanous <edtanous@google.com>
Change-Id: Id7f889de98cafaa89471d75ed3e3bb97ab3855cd
|
|
The latest version of nlohmann seems to have support for adding any
arbitrary iterable object as an array in json. Unfortunately, because
boost::urls::url produces at iterable of unsigned char, this means that
trying to encode urls leads to something like:
"@odata.id": [
47,
114,
101,
100,
102
]
Which is super unhelpful in that it does this implicitly. Given this
behavior, there are two options here, make it so that code doesn't
compile, or rely on the adl_serializer to just do the expected thing.
This patchset opts for the later, to simply to the reasonable behavior,
and call string() on the url before loading it into the json.
Tested: Unit tests passing. Fixes bug in subsequent patchset.
Signed-off-by: Ed Tanous <edtanous@google.com>
Change-Id: Id2f49bc8bd7153a0ad0c0fa8be2e13ce7c538e7f
|
|
clang-tidy has a setting, LambdaBodyIndentation, which it says:
"For callback-heavy code, it may improve readability to have the
signature indented two levels and to use OuterScope."
bmcweb is very callback heavy code. Try to enable it and see if that
improves things. There are many cases where the length of a lambda call
will change, and reindent the entire lambda function. This is really
bad for code reviews, as it's difficult to see the lines changed. This
commit should resolve it. This does have the downside of reindenting a
lot of functions, which is unfortunate, but probably worth it in the
long run.
All changes except for the .clang-format file were made by the robot.
Tested: Code compiles, whitespace changes only.
Signed-off-by: Ed Tanous <edtanous@google.com>
Change-Id: Ib4aa2f1391fada981febd25b67dcdb9143827f43
|
|
The "auth" term is overloaded in meson option and macros. This commit
changes the macro from BMCWEB_INSECURE_DISABLE_AUTHENTICATION to
BMCWEB_INSECURE_DISABLE_AUTHX, given that if "insecure-disable-auth"
is enabled, both authentication and authorization are disabled.
Tested:
1. set 'insecure-disable-auth=enabled', no authz nor authn is performed,
no crash on AccountService as well.
Signed-off-by: Nan Zhou <nanzhoumails@gmail.com>
Change-Id: Iddca1f866d16346bcc2017338fa6f077cb89cef9
|
|
Lots of #includes were missing in this file that we tangentially got
through boost/beast/websocket.hpp.
Tested: Code builds.
Signed-off-by: Ed Tanous <edtanous@google.com>
Change-Id: Iac5198f2f65eabaecf47d0fb6bb05bfa5a261f32
|
|
The existing authorization header is actually doing "authentication"
work. The authorization is happening in routing.hpp where we fetch the
role of the authenticated user and get their privilege set.
This commits changes the name of the file, as well as the namespace, to
be more precise on what the file actually does.
Tested:
1. Trivial change, it builds
Signed-off-by: Nan Zhou <nanzhoumails@gmail.com>
Change-Id: Ib91ed70507a7308522c7e5363ed2f4dc279a19d9
|
|
Modifies HttpClient so that the HTTP verb and headers can be set
for each individual message sent. Right now those fields are set
when a connection is first created and then reused by each
message sent using that connection.
Tested:
Launched two Event Listener servers that created 6 and 2
subscriptions. Sending a test event resulted in the servers
receiving 6 requests and 2 requests, respectively.
Change-Id: I8d7e2d54385bc2c403498293820adb584bff8b57
Signed-off-by: Carson Labrado <clabrado@google.com>
Signed-off-by: Ed Tanous <edtanous@google.com>
|
|
Adds sendDataWithCallback() which allows the caller to include a
callback specifying how to handle the response to that
request. This will be utilized for Redfish Aggregation
including returning the responses received when forwarding
requests to satellite BMCs.
Change-Id: I93826c8b254a5f28a982295d4145453352a90fae
Signed-off-by: Carson Labrado <clabrado@google.com>
|
|
The Redfish specification calls out that the Allow header should be
returned for all resources to give a client an indication of what
actions are allowed on that resource. The router internally has all
this data, so this patchset allows the router to construct an allow
header value, as well as return early on a HEAD request.
This was reverted once here:
https://gerrit.openbmc-project.xyz/c/openbmc/bmcweb/+/53637
Due to a redfish validator failure. With the previous patches
workaround, this error has now been resolved.
Tested:
Called curl with various parameters and observed the Allow header
curl -vvvv --insecure -X <VERB> --user root:0penBmc https://<bmc>/url
HEAD /redfish/v1/SessionService/Sessions returned Allow: GET, POST
HEAD /redfish/v1 returned Allow: GET
HEAD /redfish/v1/SessionService returned Allow: GET, PATCH
POST /redfish/v1 returned Allow: GET (method not allowed)
GET /redfish/v1 returned Allow: GET
GET /redfish/v1/SessionService returned Allow: GET, PATCH
Redfish-Protocol-Validator now reports more tests passing.
Prior to this patch:
Pass: 255, Warning: 0, Fail: 27, Not tested: 45
After this patch:
Pass: 262, Warning: 0, Fail: 21, Not tested: 43
Diff: 7 more tests passing
All tests under RESP_HEADERS_ALLOW_METHOD_NOT_ALLOWED and
RESP_HEADERS_ALLOW_GET_OR_HEAD are now passing
Included unit tests passing.
Redfish service validator is now passing.
Signed-off-by: Ed Tanous <edtanous@google.com>
Change-Id: Ibd52a7c2babe19020a0e27fa1ac79a9d33463f25
|
|
Refactors HttpClient with the following changes:
- Convert class to singleton
- Replace circular buffers with devectors
- Sending queued requests and closing connections handled
within their own callback
- Add connection pooling (max size 4)
- HttpClient supports multiple connections to multiple clients
- Retry policies can be set for specific use cases
Also modifies its use in the Subscription class to be compatible
with the refactored code.
It is assumed that a BMC will be able to handle 4 parallel
connections and thus the max pool size is set as 4. The max
number of queued messages was left unchanged at 50. Eventually
we may want to allow tuning of these limits to boost performance.
That would come in a future patch.
Tested:
Launched two Event Listener servers that created 6 and 2
subscriptions. Sending a test event created a connection pool
for each server. 4 and 2 connections were added to each pool,
respectively and were used to send the test request. For the first
pool the 2 extra requests were placed into a queue until
connections became available. After a request completed, its
associated connection was used to send the next request in
the queue. Resending the test event caused those prior connections
to be reused instead of new connections being added to the pools.
Signed-off-by: Carson Labrado <clabrado@google.com>
Change-Id: Iba72b3e342cdc05d1fb972e2e9856763a0a1b3c5
|
|
This reverts commit 867b2056d44300db9769e0d0b8883435a179834c.
Apparently we have broken the Redfish spec in a way that adding this
feature now allows the service validator to find.
@odata.id /redfish/v1/UpdateService
ERROR - Allow header should NOT contain POST for UpdateService.v1_5_0.UpdateService
Need to figure out what to do, but for now, revert to get the build
passing again.
Change-Id: Ieef20573b9caa03aba6fd2bbc999e517e4b7de3d
Signed-off-by: Ed Tanous <edtanous@google.com>
|
|
The Redfish specification calls out that the Allow header should be
returned for all resources to give a client an indication of what
actions are allowed on that resource. The router internally has all
this data, so this patchset allows the router to construct an allow
header value, as well as return early on a HEAD request.
Tested:
Called curl with various parameters and observed the Allow header
curl -vvvv --insecure -X <VERB> --user root:0penBmc https://<bmc>/url
HEAD /redfish/v1/SessionService/Sessions returned Allow: GET, POST
HEAD /redfish/v1 returned Allow: GET
HEAD /redfish/v1/SessionService returned Allow: GET, PATCH
POST /redfish/v1 returned Allow: GET (method not allowed)
GET /redfish/v1 returned Allow: GET
GET /redfish/v1/SessionService returned Allow: GET, PATCH
Redfish-Protocol-Validator now reports more tests passing.
Prior to this patch:
Pass: 255, Warning: 0, Fail: 27, Not tested: 45
After this patch:
Pass: 262, Warning: 0, Fail: 21, Not tested: 43
Diff: 7 more tests passing
All tests under RESP_HEADERS_ALLOW_METHOD_NOT_ALLOWED and
RESP_HEADERS_ALLOW_GET_OR_HEAD are now passing
Included unit tests passing.
Change-Id: Ib99835050b15eb4f419bfd21375b26e4db74fa2c
Signed-off-by: Ed Tanous <edtanous@google.com>
|
|
This commit implements a limited support for the production of etags on
json resources. It is intended to cause two things:
1. To get bmcweb to pass the PROTO_ETAG_ON_GET_ACCOUNT check, as well as
the redfish spec, which states: "Implementations shall support the
return of ETag headers for GET requests of ManagerAccount resources."
2. Begin discussions on what client-facing caching could look like in
the future, and to implement the fewest lines of code this author
could think of, with the hope of extending it later.
As written, it injects into the Response class a method that, for json
responses, uses std::hash<json> to generate an etag. This was chosen
under the assumption that it caused the least binary impact, and is
already a function provided by nlohmann, so required minimal
implementation effort to get something that functioned to the standard:
https://json.nlohmann.me/api/basic_json/std_hash/#version-history
I'm open to discussions if this should be improved in the future to
include more entropy, or to be a "weak" etag, but I think starting
with std::hash is a good first step.
This patchset intentionally does notimplement handling of the
If-None-Match, or If-Match headers that a caching client would likely
send that implements this. That is not explicitly required by the spec,
relatively complex, and probably has consequences that this author
doesn't want to write the test cases for (yet). This lack of support
makes this patchset largely only "useful" in passing the tests, and
implementing the spec to the letter, it does not generalize a caching
client feature that improves performance.
Tested:
curl -vvvv --insecure --user root:0penBmc https://192.168.7.2/redfish/v1
Contains in the response:
ETag: "765E4548"
The redfish protocol validator now passes the PROTO_ETAG_ON_GET_ACCOUNT
test, which increases our passing test count by 4 compared to previously.
Current counts are 352 passing, 30 failing, 36 not tested.
Signed-off-by: Ed Tanous <edtanous@google.com>
Change-Id: I3389b2ba98bf1276e1cb2d9c5954437b924f2d94
|
