bug#81532: Lack of software debouncing / polling rate limit causes interrupt lockup in CSM/BIOS mode on AMD B550
Pádraig Brady <[email protected]> Sat, 1 Aug 2026 16:13:11 +0100
| Newsgroups | gmane.comp.gnu.core-utils.bugs |
|---|---|
| Message-ID | <[email protected]> |
tag 81532 notabug close 81523 stop closing here as grub is not part of GNU coreutils. I suggest you resend to <[email protected]> cheers, Padraig On 01/08/2026 10:05, GC via GNU coreutils Bug Reports wrote: > > Subject: Bug: Lack of software debouncing / polling rate limit causes > interrupt lockup in CSM/BIOS mode on AMD B550 > > > > Hello GRUB Developers, > > I am reporting a severe input issue when running GRUB2 in legacy > BIOS/CSM mode on certain AMD B550 platforms (specifically the MSI MAG > B550 Tomahawk / in contrast MSI B550 MPG works fine). > > When using standard USB wired keyboards via the BIOS emulation layer, > GRUB experiences an extreme key bounce effect (every keypress is > registered 5-20 times). If a navigation key (like Arrow Down) is held > down, the entire system locks up instantly. The CPU core goes into a > hard lockup/busy-loop (power draw increases by 30W, fans spin up). > > I have tried forcing `GRUB_TERMINAL_INPUT="at_keyboard"`, but the issue > persists. It seems the BIOS emulated keyboard driver fires interrupts at > an unthrottled rate, and GRUB's real-mode/protected-mode input handler > has no built-in software debouncing or polling-rate limitation to reject > these rapid duplicate packets. > > Would it be possible to implement a basic input dumper/debouncer or a > slight delay loop inside the legacy keyboard polling routine to prevent > such hardware interrupt storms from freezing the bootloader? > > Best regards. > > > > (Mail was generated by gemini due to my bad English.) > > >