handle_qman_data_on_err(9)

πŸ“¦ version 7.3.1, in nixpkgs unstable

✨ handle_qman_data_on_err β€” extract QMAN data on error ✨

okay so bestie, welcome to page handle_qman_data_on_err(9) of the Kernel API Manual πŸ’… this one's a little diagnostic icon and you're gonna love her


🌸 SYNOPSIS 🌸

void handle_qman_data_on_err (struct hl_device *hdev , u32 qid_base ,
       u32 stream , u64 qman_base , u64 event_mask );

returns nothing, takes five args, does so much


πŸŽ€ ARGUMENTS πŸŽ€

  • hdev β€” pointer to the habanalabs device structure, like, the whole device's main character πŸ’–
  • qid_base β€” the first QID of the QMAN (out of 4 streams, yes 4, remember that number, she comes back later πŸ‘€)
  • stream β€” the QMAN's stream, literally just which stream we're talking about
  • qman_base β€” base address of the QMAN registers block, the coordinates of the vibe
  • event_mask β€” mask of the last events that occurred, AKA the receipts 🧾

πŸ¦‹ DESCRIPTION πŸ¦‹

so this function's whole job? attempt to extract as much data as possible on QMAN error 🫢 like when stuff goes wrong she is not leaving quietly, she's grabbing everything she can on the way out. no cap.

here's the tea on what actually gets printed:

  • on upper CP</> β€” print the SW config stream data and the last 8 PQEs ⭐
  • on lower CP</> β€” print SW config data and the last PQEs of ALL 4 upper CPs 😭 (told you the 4 would come back)

so yeah, upper CP keeps it to its own stream, lower CP is out here collecting from all four. generous queen πŸ‘‘


πŸ‘‘ SEE ALSO πŸ‘‘

these are her friends, go say hi πŸ’œ


and that's the whole page bestie!! brought to you by the fine folks over at ./drivers/accel/habanalabs/gaudi, dated October 2026 πŸŽ€ mwah, go read the source, ily, bye πŸ¦‹βœ¨

πŸ“¦ where this page comes from