[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] Re: [Xen-devel] [PATCH RFC 2/2] x86/HVM: split page straddling emulated accesses in more cases
> -----Original Message----- > From: Jan Beulich [mailto:JBeulich@xxxxxxxx] > Sent: 30 August 2018 12:10 > To: xen-devel <xen-devel@xxxxxxxxxxxxxxxxxxxx> > Cc: Olaf Hering <olaf@xxxxxxxxx>; Andrew Cooper > <Andrew.Cooper3@xxxxxxxxxx>; Paul Durrant <Paul.Durrant@xxxxxxxxxx>; > Wei Liu <wei.liu2@xxxxxxxxxx> > Subject: [PATCH RFC 2/2] x86/HVM: split page straddling emulated accesses > in more cases > > Assuming consecutive linear addresses map to all RAM or all MMIO is not > correct. Nor is assuming that a page straddling MMIO access will access > the same emulating component for both parts of the access. If a guest > RAM read fails with HVMTRANS_bad_gfn_to_mfn and if the access straddles > a page boundary, issue accesses separately for both parts. > > Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx> > --- > RFC: This clearly wants mirroring to the write path, and perhaps also > to the fallback code on the RMW path. But I'd like to get a sense > first on how welcome the general approach is. At first glance, of course, it appears that there would be a problem when the second linear_read() after a split returns X86EMUL_RETRY, thus causing the first one to be re-called when the emulation is re-done. Because the underlying cache handles the re-call this is not actually a problem though so I think the approach is ok and should also be fine on the write path. Paul > > --- a/xen/arch/x86/hvm/emulate.c > +++ b/xen/arch/x86/hvm/emulate.c > @@ -1041,6 +1041,48 @@ static inline int hvmemul_linear_mmio_wr > pfec, hvmemul_ctxt, translate); > } > > +static int linear_read(unsigned long addr, unsigned int bytes, void *p_data, > + uint32_t pfec, struct hvm_emulate_ctxt *hvmemul_ctxt) > +{ > + pagefault_info_t pfinfo; > + int rc = hvm_copy_from_guest_linear(p_data, addr, bytes, pfec, > &pfinfo); > + > + switch ( rc ) > + { > + unsigned int offset, part1; > + > + case HVMTRANS_okay: > + return X86EMUL_OKAY; > + > + case HVMTRANS_bad_linear_to_gfn: > + x86_emul_pagefault(pfinfo.ec, pfinfo.linear, &hvmemul_ctxt->ctxt); > + return X86EMUL_EXCEPTION; > + > + case HVMTRANS_bad_gfn_to_mfn: > + if ( pfec & PFEC_insn_fetch ) > + return X86EMUL_UNHANDLEABLE; > + > + offset = addr & ~PAGE_MASK; > + if ( offset + bytes <= PAGE_SIZE ) > + return hvmemul_linear_mmio_read(addr, bytes, p_data, pfec, > + hvmemul_ctxt, 0); > + > + /* Split the access at the page boundary. */ > + part1 = PAGE_SIZE - offset; > + rc = linear_read(addr, part1, p_data, pfec, hvmemul_ctxt); > + if ( rc == X86EMUL_OKAY ) > + rc = linear_read(addr + part1, bytes - part1, p_data + part1, > + pfec, hvmemul_ctxt); > + return rc; > + > + case HVMTRANS_gfn_paged_out: > + case HVMTRANS_gfn_shared: > + return X86EMUL_RETRY; > + } > + > + return X86EMUL_UNHANDLEABLE; > +} > + > static int __hvmemul_read( > enum x86_segment seg, > unsigned long offset, > @@ -1049,11 +1091,9 @@ static int __hvmemul_read( > enum hvm_access_type access_type, > struct hvm_emulate_ctxt *hvmemul_ctxt) > { > - struct vcpu *curr = current; > - pagefault_info_t pfinfo; > unsigned long addr, reps = 1; > uint32_t pfec = PFEC_page_present; > - struct hvm_vcpu_io *vio = &curr->arch.hvm_vcpu.hvm_io; > + struct hvm_vcpu_io *vio = ¤t->arch.hvm_vcpu.hvm_io; > int rc; > > if ( is_x86_system_segment(seg) ) > @@ -1073,28 +1113,7 @@ static int __hvmemul_read( > (vio->mmio_gla == (addr & PAGE_MASK)) ) > return hvmemul_linear_mmio_read(addr, bytes, p_data, pfec, > hvmemul_ctxt, 1); > > - rc = hvm_copy_from_guest_linear(p_data, addr, bytes, pfec, &pfinfo); > - > - switch ( rc ) > - { > - case HVMTRANS_okay: > - break; > - case HVMTRANS_bad_linear_to_gfn: > - x86_emul_pagefault(pfinfo.ec, pfinfo.linear, &hvmemul_ctxt->ctxt); > - return X86EMUL_EXCEPTION; > - case HVMTRANS_bad_gfn_to_mfn: > - if ( access_type == hvm_access_insn_fetch ) > - return X86EMUL_UNHANDLEABLE; > - > - return hvmemul_linear_mmio_read(addr, bytes, p_data, pfec, > hvmemul_ctxt, 0); > - case HVMTRANS_gfn_paged_out: > - case HVMTRANS_gfn_shared: > - return X86EMUL_RETRY; > - default: > - return X86EMUL_UNHANDLEABLE; > - } > - > - return X86EMUL_OKAY; > + return linear_read(addr, bytes, p_data, pfec, hvmemul_ctxt); > } > > static int hvmemul_read( > > > _______________________________________________ Xen-devel mailing list Xen-devel@xxxxxxxxxxxxxxxxxxxx https://lists.xenproject.org/mailman/listinfo/xen-devel
|
Lists.xenproject.org is hosted with RackSpace, monitoring our |