[CI] issue: INFINIOPS-770 Update secret-scan container registry - #1180
[CI] issue: INFINIOPS-770 Update secret-scan container registry#1180elkopel wants to merge 150 commits into
Conversation
Signed-off-by: Alexander Grissik <agrissik@nvidia.com>
Signed-off-by: Alexander Grissik <agrissik@nvidia.com>
Signed-off-by: Alexander Grissik <agrissik@nvidia.com>
Signed-off-by: Alex Briskin <abriskin@nvidia.com>
Signed-off-by: Alexander Ryabov <aryabov@nvidia.com>
…ubuntu20.04-inbox-aarch64 removed Signed-off-by: Alexander Ryabov <aryabov@nvidia.com>
Signed-off-by: Alexander Grissik <agrissik@nvidia.com>
Signed-off-by: Alexander Grissik <agrissik@nvidia.com>
Signed-off-by: Igor Ivanov <igori@nvidia.com>
Signed-off-by: Alexander Grissik <agrissik@nvidia.com>
temporarely removed Fedora 3.6 changed BlackDuck default to false Signed-off-by: Shaligin Rostislav <rshaligin@nvidia.com>
Signed-off-by: Alexander Grissik <agrissik@nvidia.com>
In reclaim buffer api - vma_socketxtreme_free_vma_packets we use reclaim_recv_buffers method which uses try_lock on the ring, but in case the lock has failed - buffers will not be returned. The fix - in case the ring is locked - reclaim buffers to the global pool. Signed-off-by: Iftah Levi <iftahl@nvidia.com>
Signed-off-by: Alexander Grissik <agrissik@nvidia.com>
Signed-off-by: Igor Ivanov <igori@nvidia.com>
Signed-off-by: Igor Ivanov <igori@nvidia.com>
See: http://0pointer.de/blog/projects/changing-roots First of all, systemctl detects when it is run in a chroot. If so, most of its operations will become NOPs, with the exception of systemctl enable and systemctl disable. If a package installation script hence calls these two commands, services will be enabled in the guest OS. However, should a package installation script include a command like systemctl restart as part of the package upgrade process this will have no effect at all when run in a chroot() environment. Signed-off-by: Igor Ivanov <igori@nvidia.com>
Equality checks in lwip on seqno related parameters must be done using special macros to a handle situations when seqno wraps around. Signed-off-by: Alexander Grissik <agrissik@nvidia.com>
With LRO/GRO, TCP segments can be large and 16bit type is not enough to fit possible offsets/lengths. This can lead to a type overflow and broken pbuf chains as result. Further, a broken chain breaks accounting of received data what leads to a warning or even a segfault. Increase type length for pbuf_header() and the trimming code. This fixes pbuf chain. Also increase type length for pbuf_realloc() to avoid similar issues in the future. Signed-off-by: Dmytro Podgornyi <dmytrop@nvidia.com>
Chain of pbufs can contain some pbufs with ref count >=1 like in ooo or flow tag flows. While processing Rx packets we may split buffer chains and we increment ref count for the new head of the chain after the split. It will cause a wrong ref count, and the buffer won't be reclaimed. Therefore we reset ref count 0 for chained mem_buf_desc in m_rx_pkt_ready_list except for the head. Related to issue 3231710. Signed-off-by: Iftah Levi <iftahl@nvidia.com>
Handle retransmitted TCP stream in GRO. In case we got a packet that its sequence number is older than what we already received, we won't aggregate it to GRO, but pass it to lwip to handle. Related to issue 3532666. Signed-off-by: Iftah Levi <iftahl@nvidia.com>
tcp_split_rexmit function used to split tcp segments by taking the pbuf chain of the orig segment and split it so the head pbuf will remain in the orig segment, and the second pbuf + its chain will be part of a new segment. The split was done incorrectly so the second pbuf had a wrong tot_len value. It wokred fine if the second pbuf has no chained pbufs. Related to issue 3291178. Signed-off-by: Iftah Levi <iftahl@nvidia.com>
Signed-off-by: Asaf Sonis <asafson@nvidia.com>
When the ring is destroyed and the RQ is empty the termination loop becomes infinte. It waits for the last_posted_wqe to be retrieved. but since it is empty the last posted wqe was already retrieved. Signed-off-by: Alexander Grissik <agrissik@nvidia.com>
Signed-off-by: Alexander Grissik <agrissik@nvidia.com>
For rx flow create key basing on allocation logic during ring creation and store it in sockinfo object. Use actual key (w/o or w/ migration sceanrio) from sockinfo object. Signed-off-by: Igor Ivanov <igori@nvidia.com>
Sockinfo object can use few different rx rings but all these rings should have the same key and identical ring allocation logic otherwise corruption happens during ring release. Signed-off-by: Igor Ivanov <igori@nvidia.com>
Ring per ip logic was introduced in "1465058 Introduce ring per ip-address logic" but was broken by "1653789 Fix Rx and Tx migrations thread safety" incorrect is_logic_support_migration() modification. As a result XLIO_RING_MIGRATION_RATIO_RX=-1 does not work. This change add special processing in ring creation/destruction. Signed-off-by: Igor Ivanov <igori@nvidia.com>
1. Enabled the feature in LwIP 2. Added support for reading the /proc/sys/net/ipv4/tcp_keepalive_* files at the XLIO startup. 3. Add support for TCP_KEEP* setsockop/getsockopt values. 4. Added unit tests. 5. Updated the CI to execute the new tests. Signed-off-by: Alex Briskin <abriskin@nvidia.com>
Set ETIMEDOUT errno and return -1 from recv in case a socket was timed out, instead of 0 return value and 0 errno. For instance, in case of TCP keep alive timeout. Signed-off-by: Alexander Grissik <agrissik@nvidia.com>
The Coverity fix in commit c4f0ae9 incorrectly checked the return value of wait_for_notification_and_process_element(). The check `if (!ret)` treats 0 as an error, but 0 is a valid return meaning "no elements processed". Return value semantics: >0 : processed N elements (success) 0 : no elements processed (normal) -1 : error, check errno EAGAIN = lock contention or notification not armed (expected) other = actual error The warning "Resource temporarily unavailable" (errno=11/EAGAIN) was flooding logs because EAGAIN is expected when the notification channel is not armed or another thread holds the lock. Fix by only warning on actual errors (ret < 0 && errno != EAGAIN), consistent with the pattern in net_device_table_mgr.cpp. Signed-off-by: Tomer Cabouly <tcabouly@nvidia.com>
Signed-off-by: Gal Noam <gnoam@nvidia.com>
After driver update the max amount of QP work elements reported by ibv_query_device is 8K instead of 32K. This limits us to create big QPs. ibv_create_qp supports up to 32K. Signed-off-by: Alexander Grissik <agrissik@nvidia.com>
Previously, VMA_BF didn’t affect BF usuage, because is_bf() checks based on MLX5_SHUT_UP_BF, and VMA_BF never adjusted it. To correctly enable/disable Blueflame, we should set MLX5_SHUT_UP_BF RDMA CORE environment variable. Signed-off-by: Bashar Abdelgafer <babdelgafer@nvidia.com>
Add a warning message when BlueFlame (BF) is requested via handle_bf but cannot be set up due to hardware or environment limitations. Signed-off-by: Bashar Abdelgafer <babdelgafer@nvidia.com>
This reverts commit d96916c.
Signed-off-by: Gal Noam <gnoam@nvidia.com>
Update DOCA version from 2.8.0 to 3.3.0. Fix doca_install.sh to use major version for DOCA repo path. Signed-off-by: Noam Tsemah <ntsemah@nvidia.com>
Signed-off-by: Noam Tsemah <ntsemah@nvidia.com>
sha1 might contain tag/<tag> as input, causing files created in the blackduck script to be under a new sub folder that does not exist Remove "tags/" prefix from sha1 Signed-off-by: NirWolfer <nwolfer@nvidia.com>
Signed-off-by: Noam Tsemah <ntsemah@nvidia.com>
Add Valgrind suppression for a sockperf memory leaks, This is a genuine sockperf leak unrelated to libvma. Signed-off-by: Bashar Abdelgafer <babdelgafer@nvidia.com>
Valgrind today is being terminated with -9 which prevents it from printing all the summaries at the end which are needed for debuging. Change -9 signal to -SIGINT af the first try of shutting down Valgrind Signed-off-by: NirWolfer <nwolfer@nvidia.com>
Name anonymous structs socket_stats_t and socket_instance_block_t to fix -Werror=non-c-typedef-for-linkage with newer compilers. Structs with member functions require a tag name for C++ linkage. Matches the existing pattern used by sh_mem_t in the same file. Signed-off-by: Thomas Cabouly <tcabouly@nvidia.com>
Build with --enable-opt-log=no compiles ring_logfunc to ((void)0), leaving 'freed' / 'count' assigned but unread and breaking -Werror=unused-but-set-variable. - ring_simple.cpp::put_tx_buffers: NOT_IN_USE(freed) before logfunc - ring_tap.cpp::mem_buf_tx_release: NOT_IN_USE(freed) before logfunc - ring_tap.cpp::mem_buf_desc_return_single_to_owner_tx: drop dead 'count' (incremented but never logged or returned) Signed-off-by: Tomer Cabouly <tcabouly@nvidia.com>
Build with --enable-opt-log=no compiles __log_funcall / __if_dbg / si_udp_logfunc to ((void)0), and the BURST variant of vma_ibv_modify_qp_rate_limit drops its mask argument, leaving variables assigned but unread and breaking -Werror=unused-but-set-variable. - verbs_extra.cpp::priv_ibv_modify_qp_ratelimit: NOT_IN_USE(attr_mask) after the modify call (mask is dropped by the macro under DEFINED_IBV_QP_SUPPORT_BURST). - io_mux_call.cpp::polling_loops: NOT_IN_USE(poll_counter) after the trailing __if_dbg (only consumer when log levels are enabled). - sockinfo_udp.cpp::rx_request_notification: NOT_IN_USE(ring_armed_count) before si_udp_logfunc (only consumer when log levels are enabled). Signed-off-by: Tomer Cabouly <tcabouly@nvidia.com>
Add CI job that builds libvma in a Fedora Rawhide container Signed-off-by: Noam Tsemah <ntsemah@nvidia.com>
📝 WalkthroughWalkthroughThe CI matrix updates the ChangesSecret Scan Image
Estimated code review effort: 1 (Trivial) | ~2 minutes Poem
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
Greptile SummaryThis PR updates the CI matrix configuration to point the secret-scan step at the new registry path (
Confidence Score: 5/5
Important Files Changed
Flowchart%%{init: {'theme': 'neutral'}}%%
flowchart TD
A[CI Matrix Job] --> B{Image Source}
B -- "Before (obsolete)" --> C["${registry_host}/toolbox/secret_scan:0.0.27"]
B -- "After (updated)" --> D["${registry_host}/swx-storage/tools/secret_scan:devops-stable"]
D --> E[Secret Scan Step]
C --> E
E --> F[CI Validation Complete]
Reviews (2): Last reviewed commit: "issue: 770 [CI] Update secret-scan conta..." | Re-trigger Greptile |
| - {name: 'blackduck', file: '.ci/dockerfiles/Dockerfile.rhel8.6', category: 'tool', arch: 'x86_64', tag: '20250630', uri: 'vma/$arch/$name/bduck', build_args: '--no-cache --target bduck'} | ||
| - {name: 'header-check', url: '${registry_host}/toolbox/header_check:0.0.58', category: 'tool', arch: 'x86_64', tag: '0.0.58'} | ||
| - {name: 'secret-scan', url: '${registry_host}/toolbox/secret_scan:0.0.27', arch: 'x86_64', tag: '0.0.27', category: 'tool'} | ||
| - {name: 'secret-scan', url: '${registry_host}/swx-storage/tools/secret_scan:devops-stable', arch: 'x86_64', tag: 'devops-stable', category: 'tool'} |
There was a problem hiding this comment.
Mutable floating tag reduces build reproducibility
The new tag devops-stable is a moving reference — any update to that image in the registry will silently change the behavior of the Secret Scan step without a corresponding change to this file. The previous pinned tag 0.0.27 ensured a specific, auditable image version was always pulled. If devops-stable is intentionally chosen by the team as a managed stable pointer, consider documenting that expectation; otherwise a pinned digest or explicit version tag would make CI behavior more predictable and easier to bisect on failures.
Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!
|
bot:rm:retest |
|
fail reason - missing redmine issue number. |
## What Use the devops-stable secret-scan image from swx-storage. ## Why ? The previous secret-scan registry path is obsolete. ## How ? Update the image URL and tag in the CI matrix.
|
Note GitHub couldn't provide a complete incremental comparison for this pull request, so CodeRabbit is performing a full review instead. This review may take a little longer. |
|
bot:retest |
INFINIOPS-770
Description
Please provide a summary of the change.
What
Use the devops-stable secret-scan image from swx-storage.
Why ?
The previous secret-scan registry path is obsolete.
How ?
Update the image URL and tag in the CI matrix.
Change type
What kind of change does this PR introduce?
Check list
Summary by CodeRabbit