Linux 7.3 delays printk kthreads so some SMP machines can finish booting

A later start for the console threads

FreshBSD's copy of the Linux tree shows a commit titled printk: Postpone kthread startup, by John Ogness via Petr Mladek, on October 16, 2026. The message says bug reports of SMP machines hanging at boot overlapped on missed inter-processor interrupts during SMP setup, and that reverting NBCON console use made the bugs disappear. It says the console-printing kthreads were started very early, from an early initcall, and that they enable and disable interrupts asynchronously while that setup is still running. Starting them later avoided the hang on machines that could reproduce it. The change waits until architecture initcalls have completed. Phoronix's October 17 story quotes that explanation and says the fix is in the Linux 7.3 line, for code that has been present since printer kthreads arrived in September 2024. The software desk isSoftware.

What an initcall level is doing in the fix

The patch is a move from one boot stage to a later one. Collabora's 2020 explainer of initcalls is not a report of this hang. It is a map of the levels, so early versus architecture is a real distinction in the kernel's boot rather than two adjectives. The 2026 commit is the change. The 2020 note is the vocabulary. A version string, and what it has to mean before a project calls a default final, is a different software question.what a version number has to mean.

The merge that carried it

The same FreshBSD page shows Linus Torvalds merging printk-for-7.3-hotfix on October 17. Petr Mladek's tag message lists three items. One is the kthread postpone, to prevent boot stalls on some hardware. Another is a switch of NBCON consoles into emergency mode during emergency restart, which is a different failure and a different brief. A typo fix is the third. Phoronix's story is about the boot hang. This section is about the postpone only. The paper isFutureweb.

Who hit the hang

The commit names Derek Barbosa as the reporter. It does not publish a list of motherboards. It says the reports vary and share a shape: SMP setup, missed IPIs, and a hang that goes away if NBCON consoles are reverted. That is enough to say the bug was real on the machines that reproduced it, and not enough to say every SMP PC in the world hung. Phoronix calls the hangs rare and varying, which matches the commit's caution. Hardware isHardware.

September 2024 is the age of the thread, not of the patch

Phoronix says printer kthreads arrived in September 2024, and that this postpone is a fix for that era of code. The patch itself is dated October 16, 2026, and the merge is October 17. A two-year-old mechanism and a two-day-old fix are both part of the story. Backports to older stable trees are something Phoronix says it presumes will follow. This brief will not invent the stable versions that have the patch until those trees say so. Security isSecurity.

Missed IPIs are a boot symptom

An inter-processor interrupt is how one CPU nudges another during startup. The commit's theory, stated as a theory, is that the console threads' interrupt activity collided with that setup on some hardware. The evidence it offers is the overlap in the reports, the disappearance of the bug when NBCON is reverted, and the disappearance of the hang when the threads start later. This page will not add a register dump. Networking, for a packet path rather than a boot hang, isNetworking.

What this page will not claim

This page will not say Linux 7.3 is released. Phoronix describes the merge as last-minute patch activity ahead of an expected 7.3 release. The merge of the tag is the event FreshBSD shows. A release announcement is a later document. This page will not say every distribution has shipped the fix. Open source, for a desktop image rather than this boot path, is Asked directly, What was starting too early? The commit says NBCON console-printing kthreads were started from an early initcall and were enabling and disabling interrupts while SMP setup was still in progress. The change waits until architecture initcalls have finished.Open Source.

A one-line move can still be the whole fix

FreshBSD's diff stat for Ogness's commit is a one-line change in kernel/printk/printk.c, plus and minus one. The size of the diff is not the size of the failure. The failure was a hang at boot. The fix is when a thread is created. Readers who equate a small diff with a small bug will misread this one. Mobile is Asked directly, Who wrote the boot-hang patch? FreshBSD shows printk: Postpone kthread startup by John Ogness, via Petr Mladek, dated October 16, 2026. The reported-by line names Derek Barbosa. Phoronix quotes the same explanation and calls the patch merged for 7.3.Mobile.

Where the patch sits

FreshBSD is the commit and the merge. Phoronix is the news account. Collabora is background on initcall levels and is not a witness to the hang. Nothing here is a workaround for a machine that is stuck, beyond what the commit itself changed in the tree. The short answer this page will stand on is this: Ogness's commit postpones printk kthread startup until after architecture initcalls, instead of an early initcall. The message says some SMP machines hung at boot with missed inter-processor interrupts, that dropping NBCON consoles made the hangs go away, and that starting the kthreads later avoided the hang on machines that could reproduce it. Linus merged that patch in the printk hotfix for 7.3.