Mikhail reported a lockdep splat related to the AMD specific ssb_state
lock:
  CPU0                       CPU1
  lock(&st->lock);
                             local_irq_disable();
                             lock(&(&sighand->siglock)->rlock);
                             lock(&st->lock);
  <Interrupt>
     lock(&(&sighand->siglock)->rlock);
  *** DEADLOCK ***
The connection between sighand->siglock and st->lock comes through seccomp,
which takes st->lock while holding sighand->siglock.
Make sure interrupts are disabled when __speculation_ctrl_update() is
invoked via prctl() -> speculation_ctrl_update(). Add a lockdep assert to
catch future offenders.
Fixes: 1f50ddb4f418 ("x86/speculation: Handle HT correctly on AMD")
Reported-by: Mikhail Gavrilov <mikhail.v.gavrilov@gmail.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Mikhail Gavrilov <mikhail.v.gavrilov@gmail.com>
Cc: Thomas Lendacky <thomas.lendacky@amd.com>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/alpine.DEB.2.21.1904141948200.4917@nanos.tec.linutronix.de
 
        u64 msr = x86_spec_ctrl_base;
        bool updmsr = false;
 
+       lockdep_assert_irqs_disabled();
+
        /*
         * If TIF_SSBD is different, select the proper mitigation
         * method. Note that if SSBD mitigation is disabled or permanentely
 
 void speculation_ctrl_update(unsigned long tif)
 {
+       unsigned long flags;
+
        /* Forced update. Make sure all relevant TIF flags are different */
-       preempt_disable();
+       local_irq_save(flags);
        __speculation_ctrl_update(~tif, tif);
-       preempt_enable();
+       local_irq_restore(flags);
 }
 
 /* Called from seccomp/prctl update */