summaryrefslogtreecommitdiff
path: root/tools/docs/parse-headers.py
diff options
context:
space:
mode:
authorPetr Oros <poros@redhat.com>2025-10-24 21:07:33 +0200
committerJakub Kicinski <kuba@kernel.org>2025-10-28 17:50:18 -0700
commit36fedc44e37e811f25666c600bce4bc027290226 (patch)
tree2a5e7c4f874a2ca7093801a77dd5c20d11491f4d /tools/docs/parse-headers.py
parent210b35d6a7ea415494ce75490c4b43b4e717d935 (diff)
dpll: fix device-id-get and pin-id-get to return errors properly
The device-id-get and pin-id-get handlers were ignoring errors from the find functions and sending empty replies instead of returning error codes to userspace. When dpll_device_find_from_nlattr() or dpll_pin_find_from_nlattr() returned an error (e.g., -EINVAL for "multiple matches" or -ENODEV for "not found"), the handlers checked `if (!IS_ERR(ptr))` and skipped adding the device/pin handle to the message, but then still sent the empty message as a successful reply. This caused userspace tools to receive empty responses with id=0 instead of proper netlink errors with extack messages like "multiple matches". The bug is visible via strace, which shows the kernel sending TWO netlink messages in response to a single request: 1. Empty reply (20 bytes, just header, no attributes): recvfrom(3, [{nlmsg_len=20, nlmsg_type=dpll, nlmsg_flags=0, ...}, {cmd=0x7, version=1}], ...) 2. NLMSG_ERROR ACK with extack (because of NLM_F_ACK flag): recvfrom(3, [{nlmsg_len=60, nlmsg_type=NLMSG_ERROR, nlmsg_flags=NLM_F_CAPPED|NLM_F_ACK_TLVS, ...}, [{error=0, msg={...}}, [{nla_type=NLMSGERR_ATTR_MSG}, "multiple matches"]]], ...) The C YNL library parses the first message, sees an empty response, and creates a result object with calloc() which zero-initializes all fields, resulting in id=0. The Python YNL library parses both messages and displays the extack from the second NLMSG_ERROR message. Fix by checking `if (IS_ERR(ptr))` first and returning the error code immediately, so that netlink properly sends only NLMSG_ERROR with the extack message to userspace. After this fix, both C and Python YNL tools receive only the NLMSG_ERROR and behave consistently. This affects: - DPLL_CMD_DEVICE_ID_GET: now properly returns error when multiple devices match the criteria (e.g., same module-name + clock-id) - DPLL_CMD_PIN_ID_GET: now properly returns error when multiple pins match the criteria (e.g., same module-name) Before fix: $ dpll pin id-get module-name ice 0 (wrong - should be error, there are 17 pins with module-name "ice") After fix: $ dpll pin id-get module-name ice Error: multiple matches (correct - kernel reports the ambiguity via extack) Fixes: 9d71b54b65b1 ("dpll: netlink: Add DPLL framework base functions") Signed-off-by: Petr Oros <poros@redhat.com> Reviewed-by: Ivan Vecera <ivecera@redhat.com> Link: https://patch.msgid.link/20251024190733.364101-1-poros@redhat.com Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Diffstat (limited to 'tools/docs/parse-headers.py')
0 files changed, 0 insertions, 0 deletions