Hi Julien, (Sending this email directly “To:” Bertrand and Stefano to catch their attention about one question that you ask them)
> On Oct 14, 2023, at 02:31, Julien Grall <[email protected]> wrote: > > Hi Henry, > > some more remarks. > > On 09/10/2023 02:03, Henry Wang wrote: >> +/* >> + * Invalidate all entries in the root page-tables. This is >> + * useful to get fault on entry and do an action. >> + * >> + * p2m_invalid_root() should not be called when the P2M is shared with >> + * the IOMMU because it will cause IOMMU fault. >> + */ >> +void p2m_invalidate_root(struct p2m_domain *p2m) > > I think this function doesn't make sense for the MPU. What would invalidate? > > The only caller outside of the file is in arch_domain_creation_finished(). > This was used for set/way emulation which I believe will be difficult to > implement efficiently for the MPU (see my other reply to this patch). > > So I would consider to move arch_domain_creation_finished() in mmu/p2m.c or > creating a new function p2m_domain_creation_finished() which is called from > arch_domain_creation_finished(). Yeah make sense, I will do the suggested change in v8. > [...] > >> +/* >> + * Clean & invalidate RAM associated to the guest vCPU. >> + * >> + * The function can only work with the current vCPU and should be called >> + * with IRQ enabled as the vCPU could get preempted. >> + */ >> +void p2m_flush_vm(struct vcpu *v) > I believe an implementation of p2m_flush_vm() would be quite expensive for > the MPU. It would be better to not emulate set/way for now. > > It would also feel a bit odd to leave it unimplemented and called from > check_for_vcpu_work(). So maybe we want to rename it. I don't have a goood > name suggestion though. Bertrand, Stefano? See the beginning of this email. Kind regards, Henry > > [...] > > -- > Julien Grall
