Re: [PATCH] target/arm: Fix BTI versus CF_PCREL

[Richard Henderson]

On 7/30/24 19:30, Peter Maydell wrote:
> > +static bool is_guarded_page(CPUARMState *env, target_ulong addr)
> > +{
> > +#ifdef CONFIG_USER_ONLY
> > +    return page_get_flags(addr) & PAGE_BTI;
> > +#else
> > +    CPUTLBEntryFull *full;
> > +    void *host;
> > +    int mmu_idx = cpu_mmu_index(env_cpu(env), true);
> > +    int flags = probe_access_full(env, addr, 0, MMU_INST_FETCH, mmu_idx,
> > +                                  false, &host, &full, 0);
> > +
> > +    assert(!(flags & TLB_INVALID_MASK));
>
> Is there a race condition here where some other vCPU
> knocks this entry out of the TLB between the point when
> we started executing the TB and when we made this helper
> function call ?

I don't think so, because cross-cpu flushes use async_safe_run_on_cpu, which will wait 
until this cpu returns to the main loop.  But it's just as easy to allow this probe to 
fault and unwind, Just In Case.


r~