)]}'
{
  "commit": "0df0bcfc7ccce132d1f28838153588765351a827",
  "tree": "648fb92ff4778c6f598dfc5f797231634519a402",
  "parents": [
    "40afef7c3f24c04fc75e2be632fbad1adecb2128"
  ],
  "author": {
    "name": "Ard Biesheuvel",
    "email": "ardb@kernel.org",
    "time": "Wed May 14 17:55:08 2025 +0200"
  },
  "committer": {
    "name": "Ard Biesheuvel",
    "email": "ardb@kernel.org",
    "time": "Fri Sep 05 14:55:24 2025 +0200"
  },
  "message": "arm64/efi: Call EFI runtime services without disabling preemption\n\nThe only remaining reason why EFI runtime services are invoked with\npreemption disabled is the fact that the mm is swapped out behind the\nback of the context switching code.\n\nThe kernel no longer disables preemption in kernel_neon_begin().\nFurthermore, the EFI spec is being clarified to explicitly state that\nonly baseline FP/SIMD is permitted in EFI runtime service\nimplementations, and so the existing kernel mode NEON context switching\ncode is sufficient to preserve and restore the execution context of an\nin-progress EFI runtime service call.\n\nMost EFI calls are made from the efi_rts_wq, which is serviced by a\nkthread. As kthreads never return to user space, they usually don\u0027t have\nan mm, and so we can use the existing infrastructure to swap in the\nefi_mm while the EFI call is in progress. This is visible to the\nscheduler, which will therefore reactivate the selected mm when\nswitching out the kthread and back in again.\n\nGiven that the EFI spec explicitly permits runtime services to be called\nwith interrupts enabled, firmware code is already required to tolerate\ninterruptions. So rather than disable preemption, disable only migration\nso that EFI runtime services are less likely to cause scheduling delays.\nTo avoid potential issues where runtime services are interrupted while\npolling the secure firmware for async completions, keep migration\ndisabled so that a runtime service invocation does not resume on a\ndifferent CPU from the one it was started on.\n\nNote, though, that the firmware executes at the same privilege level as\nthe kernel, and is therefore able to disable interrupts altogether.\n\nSigned-off-by: Ard Biesheuvel \u003cardb@kernel.org\u003e\n",
  "tree_diff": [
    {
      "type": "modify",
      "old_id": "ece046bcf0db40d462be7f46bd82283a3c13ceb6",
      "old_mode": 33188,
      "old_path": "arch/arm64/kernel/efi.c",
      "new_id": "cf62980006ea7151b2c38ac4828ef746d0b8d78f",
      "new_mode": 33188,
      "new_path": "arch/arm64/kernel/efi.c"
    }
  ]
}