Jeremy Fitzhardinge
2009-Jan-28 22:35 UTC
[Xen-devel] [PATCH 0 of 7] x86/paravirt: optimise pvop calls and register use
Hi Ingo,
This series implements a sequence of optimisations to reduce the
impact of enabling CONFIG_PARAVIRT while running native.
They are:
(0. Move some Xen code around to make later changes work better.)
1. For a number of the pvops, the native implemention is a simple
identity function which returns its argument. Add a specific
paravirt identity function which the patcher can treat specially,
by directly inlining the either nops (32-bit) or a mov (64-bit)
into the instruction stream.
2. When a pvop is called from asm code, it also provides a hint
about what registers are available to be clobbered by the called
code. Until now, that information was ignored, and all
caller-save registers were saved. Now, don''t bother
saving/restoring registers which are clobberable.
3. The C calling convention lists which registers the caller can
expect to survive a function call, and which the callee is
allowed to clobber. The latter set is quite large, especially on
64-bit. This means that converting a pile of simple inline
functions into function calls caused a lot more register
pressure, making the generated code much worse.
I introduce a new "callee-save" calling convention which makes
only the return register (eax:edx on 32-bit, rax on 64)
callee-clobberable; the callee must preserve all other registers,
including the argument registers.
This makes the callsites for these functions clobber many fewer
registers, giving the compiler a chance to generate better code.
Small asm functions, which generally only use one or two
registers anyway, to be directly called. C code can also be
called via a thunk, which does the necessary register
saving/restoring (generated by PV_CALLEE_SAVE_REGS_THUNK(func)).
The irq enable/disable/save/restore functions are the first to
make use of this calling convention, since they are the most
commonly used in the kernel, and are also called form asm code.
4. Convert the pte_val/make_pte identity functions to use the
callee-save convention; they''re only identity functions anyway,
so they have no need to trash lots of registers.
I had to make some adjustments to VSMP and lguest to match the new
calling conventions. I wasn''t sure how I should change VMI, so
I''m
waiting for Zach''s input on that (VMI doesn''t compile at the
moment).
In testing, the net result was that the overhead dropped by about 75%,
though I found it hard to really get stable results. The most obvious
improvement was a reduction in L2 references, presumably meaning that
L1 was getting a better hit rate. Each of these transforms is an
unambigious improvement in generated code for the native case, so I''m
curious to see what other people see.
Thanks,
J
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel
Jeremy Fitzhardinge
2009-Jan-28 22:35 UTC
[Xen-devel] [PATCH 1 of 7] xen: move remaining mmu-related stuff into mmu.c
A general cleanup, and lay the groundwork for later patches.
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
---
arch/x86/xen/enlighten.c | 731 ----------------------------------------------
arch/x86/xen/mmu.c | 700 ++++++++++++++++++++++++++++++++++++++++++++
arch/x86/xen/mmu.h | 3
arch/x86/xen/xen-ops.h | 8
4 files changed, 712 insertions(+), 730 deletions(-)
diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c
--- a/arch/x86/xen/enlighten.c
+++ b/arch/x86/xen/enlighten.c
@@ -61,35 +61,6 @@
enum xen_domain_type xen_domain_type = XEN_NATIVE;
EXPORT_SYMBOL_GPL(xen_domain_type);
-/*
- * Identity map, in addition to plain kernel map. This needs to be
- * large enough to allocate page table pages to allocate the rest.
- * Each page can map 2MB.
- */
-static pte_t level1_ident_pgt[PTRS_PER_PTE * 4] __page_aligned_bss;
-
-#ifdef CONFIG_X86_64
-/* l3 pud for userspace vsyscall mapping */
-static pud_t level3_user_vsyscall[PTRS_PER_PUD] __page_aligned_bss;
-#endif /* CONFIG_X86_64 */
-
-/*
- * Note about cr3 (pagetable base) values:
- *
- * xen_cr3 contains the current logical cr3 value; it contains the
- * last set cr3. This may not be the current effective cr3, because
- * its update may be being lazily deferred. However, a vcpu looking
- * at its own cr3 can use this value knowing that it everything will
- * be self-consistent.
- *
- * xen_current_cr3 contains the actual vcpu cr3; it is set once the
- * hypercall to set the vcpu cr3 is complete (so it may be a little
- * out of date, but it will never be set early). If one vcpu is
- * looking at another vcpu''s cr3 value, it should use this variable.
- */
-DEFINE_PER_CPU(unsigned long, xen_cr3); /* cr3 stored as physaddr */
-DEFINE_PER_CPU(unsigned long, xen_current_cr3); /* actual vcpu cr3 */
-
struct start_info *xen_start_info;
EXPORT_SYMBOL_GPL(xen_start_info);
@@ -237,7 +208,7 @@
return HYPERVISOR_get_debugreg(reg);
}
-static void xen_leave_lazy(void)
+void xen_leave_lazy(void)
{
paravirt_leave_lazy(paravirt_get_lazy_mode());
xen_mc_flush();
@@ -598,76 +569,6 @@
#endif
-static void xen_flush_tlb(void)
-{
- struct mmuext_op *op;
- struct multicall_space mcs;
-
- preempt_disable();
-
- mcs = xen_mc_entry(sizeof(*op));
-
- op = mcs.args;
- op->cmd = MMUEXT_TLB_FLUSH_LOCAL;
- MULTI_mmuext_op(mcs.mc, op, 1, NULL, DOMID_SELF);
-
- xen_mc_issue(PARAVIRT_LAZY_MMU);
-
- preempt_enable();
-}
-
-static void xen_flush_tlb_single(unsigned long addr)
-{
- struct mmuext_op *op;
- struct multicall_space mcs;
-
- preempt_disable();
-
- mcs = xen_mc_entry(sizeof(*op));
- op = mcs.args;
- op->cmd = MMUEXT_INVLPG_LOCAL;
- op->arg1.linear_addr = addr & PAGE_MASK;
- MULTI_mmuext_op(mcs.mc, op, 1, NULL, DOMID_SELF);
-
- xen_mc_issue(PARAVIRT_LAZY_MMU);
-
- preempt_enable();
-}
-
-static void xen_flush_tlb_others(const struct cpumask *cpus,
- struct mm_struct *mm, unsigned long va)
-{
- struct {
- struct mmuext_op op;
- DECLARE_BITMAP(mask, NR_CPUS);
- } *args;
- struct multicall_space mcs;
-
- BUG_ON(cpumask_empty(cpus));
- BUG_ON(!mm);
-
- mcs = xen_mc_entry(sizeof(*args));
- args = mcs.args;
- args->op.arg2.vcpumask = to_cpumask(args->mask);
-
- /* Remove us, and any offline CPUS. */
- cpumask_and(to_cpumask(args->mask), cpus, cpu_online_mask);
- cpumask_clear_cpu(smp_processor_id(), to_cpumask(args->mask));
- if (unlikely(cpumask_empty(to_cpumask(args->mask))))
- goto issue;
-
- if (va == TLB_FLUSH_ALL) {
- args->op.cmd = MMUEXT_TLB_FLUSH_MULTI;
- } else {
- args->op.cmd = MMUEXT_INVLPG_MULTI;
- args->op.arg1.linear_addr = va;
- }
-
- MULTI_mmuext_op(mcs.mc, &args->op, 1, NULL, DOMID_SELF);
-
-issue:
- xen_mc_issue(PARAVIRT_LAZY_MMU);
-}
static void xen_clts(void)
{
@@ -693,21 +594,6 @@
xen_mc_issue(PARAVIRT_LAZY_CPU);
}
-static void xen_write_cr2(unsigned long cr2)
-{
- percpu_read(xen_vcpu)->arch.cr2 = cr2;
-}
-
-static unsigned long xen_read_cr2(void)
-{
- return percpu_read(xen_vcpu)->arch.cr2;
-}
-
-static unsigned long xen_read_cr2_direct(void)
-{
- return percpu_read(xen_vcpu_info.arch.cr2);
-}
-
static void xen_write_cr4(unsigned long cr4)
{
cr4 &= ~X86_CR4_PGE;
@@ -716,71 +602,6 @@
native_write_cr4(cr4);
}
-static unsigned long xen_read_cr3(void)
-{
- return percpu_read(xen_cr3);
-}
-
-static void set_current_cr3(void *v)
-{
- percpu_write(xen_current_cr3, (unsigned long)v);
-}
-
-static void __xen_write_cr3(bool kernel, unsigned long cr3)
-{
- struct mmuext_op *op;
- struct multicall_space mcs;
- unsigned long mfn;
-
- if (cr3)
- mfn = pfn_to_mfn(PFN_DOWN(cr3));
- else
- mfn = 0;
-
- WARN_ON(mfn == 0 && kernel);
-
- mcs = __xen_mc_entry(sizeof(*op));
-
- op = mcs.args;
- op->cmd = kernel ? MMUEXT_NEW_BASEPTR : MMUEXT_NEW_USER_BASEPTR;
- op->arg1.mfn = mfn;
-
- MULTI_mmuext_op(mcs.mc, op, 1, NULL, DOMID_SELF);
-
- if (kernel) {
- percpu_write(xen_cr3, cr3);
-
- /* Update xen_current_cr3 once the batch has actually
- been submitted. */
- xen_mc_callback(set_current_cr3, (void *)cr3);
- }
-}
-
-static void xen_write_cr3(unsigned long cr3)
-{
- BUG_ON(preemptible());
-
- xen_mc_batch(); /* disables interrupts */
-
- /* Update while interrupts are disabled, so its atomic with
- respect to ipis */
- percpu_write(xen_cr3, cr3);
-
- __xen_write_cr3(true, cr3);
-
-#ifdef CONFIG_X86_64
- {
- pgd_t *user_pgd = xen_get_user_pgd(__va(cr3));
- if (user_pgd)
- __xen_write_cr3(false, __pa(user_pgd));
- else
- __xen_write_cr3(false, 0);
- }
-#endif
-
- xen_mc_issue(PARAVIRT_LAZY_CPU); /* interrupts restored */
-}
-
static int xen_write_msr_safe(unsigned int msr, unsigned low, unsigned high)
{
int ret;
@@ -822,185 +643,6 @@
return ret;
}
-/* Early in boot, while setting up the initial pagetable, assume
- everything is pinned. */
-static __init void xen_alloc_pte_init(struct mm_struct *mm, unsigned long pfn)
-{
-#ifdef CONFIG_FLATMEM
- BUG_ON(mem_map); /* should only be used early */
-#endif
- make_lowmem_page_readonly(__va(PFN_PHYS(pfn)));
-}
-
-/* Early release_pte assumes that all pts are pinned, since there''s
- only init_mm and anything attached to that is pinned. */
-static void xen_release_pte_init(unsigned long pfn)
-{
- make_lowmem_page_readwrite(__va(PFN_PHYS(pfn)));
-}
-
-static void pin_pagetable_pfn(unsigned cmd, unsigned long pfn)
-{
- struct mmuext_op op;
- op.cmd = cmd;
- op.arg1.mfn = pfn_to_mfn(pfn);
- if (HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF))
- BUG();
-}
-
-/* This needs to make sure the new pte page is pinned iff its being
- attached to a pinned pagetable. */
-static void xen_alloc_ptpage(struct mm_struct *mm, unsigned long pfn, unsigned
level)
-{
- struct page *page = pfn_to_page(pfn);
-
- if (PagePinned(virt_to_page(mm->pgd))) {
- SetPagePinned(page);
-
- vm_unmap_aliases();
- if (!PageHighMem(page)) {
- make_lowmem_page_readonly(__va(PFN_PHYS((unsigned long)pfn)));
- if (level == PT_PTE && USE_SPLIT_PTLOCKS)
- pin_pagetable_pfn(MMUEXT_PIN_L1_TABLE, pfn);
- } else {
- /* make sure there are no stray mappings of
- this page */
- kmap_flush_unused();
- }
- }
-}
-
-static void xen_alloc_pte(struct mm_struct *mm, unsigned long pfn)
-{
- xen_alloc_ptpage(mm, pfn, PT_PTE);
-}
-
-static void xen_alloc_pmd(struct mm_struct *mm, unsigned long pfn)
-{
- xen_alloc_ptpage(mm, pfn, PT_PMD);
-}
-
-static int xen_pgd_alloc(struct mm_struct *mm)
-{
- pgd_t *pgd = mm->pgd;
- int ret = 0;
-
- BUG_ON(PagePinned(virt_to_page(pgd)));
-
-#ifdef CONFIG_X86_64
- {
- struct page *page = virt_to_page(pgd);
- pgd_t *user_pgd;
-
- BUG_ON(page->private != 0);
-
- ret = -ENOMEM;
-
- user_pgd = (pgd_t *)__get_free_page(GFP_KERNEL | __GFP_ZERO);
- page->private = (unsigned long)user_pgd;
-
- if (user_pgd != NULL) {
- user_pgd[pgd_index(VSYSCALL_START)] - __pgd(__pa(level3_user_vsyscall) |
_PAGE_TABLE);
- ret = 0;
- }
-
- BUG_ON(PagePinned(virt_to_page(xen_get_user_pgd(pgd))));
- }
-#endif
-
- return ret;
-}
-
-static void xen_pgd_free(struct mm_struct *mm, pgd_t *pgd)
-{
-#ifdef CONFIG_X86_64
- pgd_t *user_pgd = xen_get_user_pgd(pgd);
-
- if (user_pgd)
- free_page((unsigned long)user_pgd);
-#endif
-}
-
-/* This should never happen until we''re OK to use struct page */
-static void xen_release_ptpage(unsigned long pfn, unsigned level)
-{
- struct page *page = pfn_to_page(pfn);
-
- if (PagePinned(page)) {
- if (!PageHighMem(page)) {
- if (level == PT_PTE && USE_SPLIT_PTLOCKS)
- pin_pagetable_pfn(MMUEXT_UNPIN_TABLE, pfn);
- make_lowmem_page_readwrite(__va(PFN_PHYS(pfn)));
- }
- ClearPagePinned(page);
- }
-}
-
-static void xen_release_pte(unsigned long pfn)
-{
- xen_release_ptpage(pfn, PT_PTE);
-}
-
-static void xen_release_pmd(unsigned long pfn)
-{
- xen_release_ptpage(pfn, PT_PMD);
-}
-
-#if PAGETABLE_LEVELS == 4
-static void xen_alloc_pud(struct mm_struct *mm, unsigned long pfn)
-{
- xen_alloc_ptpage(mm, pfn, PT_PUD);
-}
-
-static void xen_release_pud(unsigned long pfn)
-{
- xen_release_ptpage(pfn, PT_PUD);
-}
-#endif
-
-#ifdef CONFIG_HIGHPTE
-static void *xen_kmap_atomic_pte(struct page *page, enum km_type type)
-{
- pgprot_t prot = PAGE_KERNEL;
-
- if (PagePinned(page))
- prot = PAGE_KERNEL_RO;
-
- if (0 && PageHighMem(page))
- printk("mapping highpte %lx type %d prot %s\n",
- page_to_pfn(page), type,
- (unsigned long)pgprot_val(prot) & _PAGE_RW ? "WRITE" :
"READ");
-
- return kmap_atomic_prot(page, type, prot);
-}
-#endif
-
-#ifdef CONFIG_X86_32
-static __init pte_t mask_rw_pte(pte_t *ptep, pte_t pte)
-{
- /* If there''s an existing pte, then don''t allow _PAGE_RW to
be set */
- if (pte_val_ma(*ptep) & _PAGE_PRESENT)
- pte = __pte_ma(((pte_val_ma(*ptep) & _PAGE_RW) | ~_PAGE_RW) &
- pte_val_ma(pte));
-
- return pte;
-}
-
-/* Init-time set_pte while constructing initial pagetables, which
- doesn''t allow RO pagetable pages to be remapped RW */
-static __init void xen_set_pte_init(pte_t *ptep, pte_t pte)
-{
- pte = mask_rw_pte(ptep, pte);
-
- xen_set_pte(ptep, pte);
-}
-#endif
-
-static __init void xen_pagetable_setup_start(pgd_t *base)
-{
-}
-
void xen_setup_shared_info(void)
{
if (!xen_feature(XENFEAT_auto_translated_physmap)) {
@@ -1021,37 +663,6 @@
xen_setup_mfn_list_list();
}
-static __init void xen_pagetable_setup_done(pgd_t *base)
-{
- xen_setup_shared_info();
-}
-
-static __init void xen_post_allocator_init(void)
-{
- pv_mmu_ops.set_pte = xen_set_pte;
- pv_mmu_ops.set_pmd = xen_set_pmd;
- pv_mmu_ops.set_pud = xen_set_pud;
-#if PAGETABLE_LEVELS == 4
- pv_mmu_ops.set_pgd = xen_set_pgd;
-#endif
-
- /* This will work as long as patching hasn''t happened yet
- (which it hasn''t) */
- pv_mmu_ops.alloc_pte = xen_alloc_pte;
- pv_mmu_ops.alloc_pmd = xen_alloc_pmd;
- pv_mmu_ops.release_pte = xen_release_pte;
- pv_mmu_ops.release_pmd = xen_release_pmd;
-#if PAGETABLE_LEVELS == 4
- pv_mmu_ops.alloc_pud = xen_alloc_pud;
- pv_mmu_ops.release_pud = xen_release_pud;
-#endif
-
-#ifdef CONFIG_X86_64
- SetPagePinned(virt_to_page(level3_user_vsyscall));
-#endif
- xen_mark_init_mm_pinned();
-}
-
/* This is called once we have the cpu_possible_map */
void xen_setup_vcpu_info_placement(void)
{
@@ -1126,49 +737,6 @@
return ret;
}
-static void xen_set_fixmap(unsigned idx, unsigned long phys, pgprot_t prot)
-{
- pte_t pte;
-
- phys >>= PAGE_SHIFT;
-
- switch (idx) {
- case FIX_BTMAP_END ... FIX_BTMAP_BEGIN:
-#ifdef CONFIG_X86_F00F_BUG
- case FIX_F00F_IDT:
-#endif
-#ifdef CONFIG_X86_32
- case FIX_WP_TEST:
- case FIX_VDSO:
-# ifdef CONFIG_HIGHMEM
- case FIX_KMAP_BEGIN ... FIX_KMAP_END:
-# endif
-#else
- case VSYSCALL_LAST_PAGE ... VSYSCALL_FIRST_PAGE:
-#endif
-#ifdef CONFIG_X86_LOCAL_APIC
- case FIX_APIC_BASE: /* maps dummy local APIC */
-#endif
- pte = pfn_pte(phys, prot);
- break;
-
- default:
- pte = mfn_pte(phys, prot);
- break;
- }
-
- __native_set_fixmap(idx, pte);
-
-#ifdef CONFIG_X86_64
- /* Replicate changes to map the vsyscall page into the user
- pagetable vsyscall mapping. */
- if (idx >= VSYSCALL_LAST_PAGE && idx <= VSYSCALL_FIRST_PAGE) {
- unsigned long vaddr = __fix_to_virt(idx);
- set_pte_vaddr_pud(level3_user_vsyscall, vaddr, pte);
- }
-#endif
-}
-
static const struct pv_info xen_info __initdata = {
.paravirt_enabled = 1,
.shared_kernel_pmd = 0,
@@ -1264,86 +832,6 @@
#endif
};
-static const struct pv_mmu_ops xen_mmu_ops __initdata = {
- .pagetable_setup_start = xen_pagetable_setup_start,
- .pagetable_setup_done = xen_pagetable_setup_done,
-
- .read_cr2 = xen_read_cr2,
- .write_cr2 = xen_write_cr2,
-
- .read_cr3 = xen_read_cr3,
- .write_cr3 = xen_write_cr3,
-
- .flush_tlb_user = xen_flush_tlb,
- .flush_tlb_kernel = xen_flush_tlb,
- .flush_tlb_single = xen_flush_tlb_single,
- .flush_tlb_others = xen_flush_tlb_others,
-
- .pte_update = paravirt_nop,
- .pte_update_defer = paravirt_nop,
-
- .pgd_alloc = xen_pgd_alloc,
- .pgd_free = xen_pgd_free,
-
- .alloc_pte = xen_alloc_pte_init,
- .release_pte = xen_release_pte_init,
- .alloc_pmd = xen_alloc_pte_init,
- .alloc_pmd_clone = paravirt_nop,
- .release_pmd = xen_release_pte_init,
-
-#ifdef CONFIG_HIGHPTE
- .kmap_atomic_pte = xen_kmap_atomic_pte,
-#endif
-
-#ifdef CONFIG_X86_64
- .set_pte = xen_set_pte,
-#else
- .set_pte = xen_set_pte_init,
-#endif
- .set_pte_at = xen_set_pte_at,
- .set_pmd = xen_set_pmd_hyper,
-
- .ptep_modify_prot_start = __ptep_modify_prot_start,
- .ptep_modify_prot_commit = __ptep_modify_prot_commit,
-
- .pte_val = xen_pte_val,
- .pgd_val = xen_pgd_val,
-
- .make_pte = xen_make_pte,
- .make_pgd = xen_make_pgd,
-
-#ifdef CONFIG_X86_PAE
- .set_pte_atomic = xen_set_pte_atomic,
- .set_pte_present = xen_set_pte_at,
- .pte_clear = xen_pte_clear,
- .pmd_clear = xen_pmd_clear,
-#endif /* CONFIG_X86_PAE */
- .set_pud = xen_set_pud_hyper,
-
- .make_pmd = xen_make_pmd,
- .pmd_val = xen_pmd_val,
-
-#if PAGETABLE_LEVELS == 4
- .pud_val = xen_pud_val,
- .make_pud = xen_make_pud,
- .set_pgd = xen_set_pgd_hyper,
-
- .alloc_pud = xen_alloc_pte_init,
- .release_pud = xen_release_pte_init,
-#endif /* PAGETABLE_LEVELS == 4 */
-
- .activate_mm = xen_activate_mm,
- .dup_mmap = xen_dup_mmap,
- .exit_mmap = xen_exit_mmap,
-
- .lazy_mode = {
- .enter = paravirt_enter_lazy_mmu,
- .leave = xen_leave_lazy,
- },
-
- .set_fixmap = xen_set_fixmap,
-};
-
static void xen_reboot(int reason)
{
struct sched_shutdown r = { .reason = reason };
@@ -1386,223 +874,6 @@
};
-static void __init xen_reserve_top(void)
-{
-#ifdef CONFIG_X86_32
- unsigned long top = HYPERVISOR_VIRT_START;
- struct xen_platform_parameters pp;
-
- if (HYPERVISOR_xen_version(XENVER_platform_parameters, &pp) == 0)
- top = pp.virt_start;
-
- reserve_top_address(-top);
-#endif /* CONFIG_X86_32 */
-}
-
-/*
- * Like __va(), but returns address in the kernel mapping (which is
- * all we have until the physical memory mapping has been set up.
- */
-static void *__ka(phys_addr_t paddr)
-{
-#ifdef CONFIG_X86_64
- return (void *)(paddr + __START_KERNEL_map);
-#else
- return __va(paddr);
-#endif
-}
-
-/* Convert a machine address to physical address */
-static unsigned long m2p(phys_addr_t maddr)
-{
- phys_addr_t paddr;
-
- maddr &= PTE_PFN_MASK;
- paddr = mfn_to_pfn(maddr >> PAGE_SHIFT) << PAGE_SHIFT;
-
- return paddr;
-}
-
-/* Convert a machine address to kernel virtual */
-static void *m2v(phys_addr_t maddr)
-{
- return __ka(m2p(maddr));
-}
-
-static void set_page_prot(void *addr, pgprot_t prot)
-{
- unsigned long pfn = __pa(addr) >> PAGE_SHIFT;
- pte_t pte = pfn_pte(pfn, prot);
-
- if (HYPERVISOR_update_va_mapping((unsigned long)addr, pte, 0))
- BUG();
-}
-
-static __init void xen_map_identity_early(pmd_t *pmd, unsigned long max_pfn)
-{
- unsigned pmdidx, pteidx;
- unsigned ident_pte;
- unsigned long pfn;
-
- ident_pte = 0;
- pfn = 0;
- for (pmdidx = 0; pmdidx < PTRS_PER_PMD && pfn < max_pfn;
pmdidx++) {
- pte_t *pte_page;
-
- /* Reuse or allocate a page of ptes */
- if (pmd_present(pmd[pmdidx]))
- pte_page = m2v(pmd[pmdidx].pmd);
- else {
- /* Check for free pte pages */
- if (ident_pte == ARRAY_SIZE(level1_ident_pgt))
- break;
-
- pte_page = &level1_ident_pgt[ident_pte];
- ident_pte += PTRS_PER_PTE;
-
- pmd[pmdidx] = __pmd(__pa(pte_page) | _PAGE_TABLE);
- }
-
- /* Install mappings */
- for (pteidx = 0; pteidx < PTRS_PER_PTE; pteidx++, pfn++) {
- pte_t pte;
-
- if (pfn > max_pfn_mapped)
- max_pfn_mapped = pfn;
-
- if (!pte_none(pte_page[pteidx]))
- continue;
-
- pte = pfn_pte(pfn, PAGE_KERNEL_EXEC);
- pte_page[pteidx] = pte;
- }
- }
-
- for (pteidx = 0; pteidx < ident_pte; pteidx += PTRS_PER_PTE)
- set_page_prot(&level1_ident_pgt[pteidx], PAGE_KERNEL_RO);
-
- set_page_prot(pmd, PAGE_KERNEL_RO);
-}
-
-#ifdef CONFIG_X86_64
-static void convert_pfn_mfn(void *v)
-{
- pte_t *pte = v;
- int i;
-
- /* All levels are converted the same way, so just treat them
- as ptes. */
- for (i = 0; i < PTRS_PER_PTE; i++)
- pte[i] = xen_make_pte(pte[i].pte);
-}
-
-/*
- * Set up the inital kernel pagetable.
- *
- * We can construct this by grafting the Xen provided pagetable into
- * head_64.S''s preconstructed pagetables. We copy the Xen
L2''s into
- * level2_ident_pgt, level2_kernel_pgt and level2_fixmap_pgt. This
- * means that only the kernel has a physical mapping to start with -
- * but that''s enough to get __va working. We need to fill in the rest
- * of the physical mapping once some sort of allocator has been set
- * up.
- */
-static __init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd,
- unsigned long max_pfn)
-{
- pud_t *l3;
- pmd_t *l2;
-
- /* Zap identity mapping */
- init_level4_pgt[0] = __pgd(0);
-
- /* Pre-constructed entries are in pfn, so convert to mfn */
- convert_pfn_mfn(init_level4_pgt);
- convert_pfn_mfn(level3_ident_pgt);
- convert_pfn_mfn(level3_kernel_pgt);
-
- l3 = m2v(pgd[pgd_index(__START_KERNEL_map)].pgd);
- l2 = m2v(l3[pud_index(__START_KERNEL_map)].pud);
-
- memcpy(level2_ident_pgt, l2, sizeof(pmd_t) * PTRS_PER_PMD);
- memcpy(level2_kernel_pgt, l2, sizeof(pmd_t) * PTRS_PER_PMD);
-
- l3 = m2v(pgd[pgd_index(__START_KERNEL_map + PMD_SIZE)].pgd);
- l2 = m2v(l3[pud_index(__START_KERNEL_map + PMD_SIZE)].pud);
- memcpy(level2_fixmap_pgt, l2, sizeof(pmd_t) * PTRS_PER_PMD);
-
- /* Set up identity map */
- xen_map_identity_early(level2_ident_pgt, max_pfn);
-
- /* Make pagetable pieces RO */
- set_page_prot(init_level4_pgt, PAGE_KERNEL_RO);
- set_page_prot(level3_ident_pgt, PAGE_KERNEL_RO);
- set_page_prot(level3_kernel_pgt, PAGE_KERNEL_RO);
- set_page_prot(level3_user_vsyscall, PAGE_KERNEL_RO);
- set_page_prot(level2_kernel_pgt, PAGE_KERNEL_RO);
- set_page_prot(level2_fixmap_pgt, PAGE_KERNEL_RO);
-
- /* Pin down new L4 */
- pin_pagetable_pfn(MMUEXT_PIN_L4_TABLE,
- PFN_DOWN(__pa_symbol(init_level4_pgt)));
-
- /* Unpin Xen-provided one */
- pin_pagetable_pfn(MMUEXT_UNPIN_TABLE, PFN_DOWN(__pa(pgd)));
-
- /* Switch over */
- pgd = init_level4_pgt;
-
- /*
- * At this stage there can be no user pgd, and no page
- * structure to attach it to, so make sure we just set kernel
- * pgd.
- */
- xen_mc_batch();
- __xen_write_cr3(true, __pa(pgd));
- xen_mc_issue(PARAVIRT_LAZY_CPU);
-
- reserve_early(__pa(xen_start_info->pt_base),
- __pa(xen_start_info->pt_base +
- xen_start_info->nr_pt_frames * PAGE_SIZE),
- "XEN PAGETABLES");
-
- return pgd;
-}
-#else /* !CONFIG_X86_64 */
-static pmd_t level2_kernel_pgt[PTRS_PER_PMD] __page_aligned_bss;
-
-static __init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd,
- unsigned long max_pfn)
-{
- pmd_t *kernel_pmd;
-
- init_pg_tables_start = __pa(pgd);
- init_pg_tables_end = __pa(pgd) + xen_start_info->nr_pt_frames*PAGE_SIZE;
- max_pfn_mapped = PFN_DOWN(init_pg_tables_end + 512*1024);
-
- kernel_pmd = m2v(pgd[KERNEL_PGD_BOUNDARY].pgd);
- memcpy(level2_kernel_pgt, kernel_pmd, sizeof(pmd_t) * PTRS_PER_PMD);
-
- xen_map_identity_early(level2_kernel_pgt, max_pfn);
-
- memcpy(swapper_pg_dir, pgd, sizeof(pgd_t) * PTRS_PER_PGD);
- set_pgd(&swapper_pg_dir[KERNEL_PGD_BOUNDARY],
- __pgd(__pa(level2_kernel_pgt) | _PAGE_PRESENT));
-
- set_page_prot(level2_kernel_pgt, PAGE_KERNEL_RO);
- set_page_prot(swapper_pg_dir, PAGE_KERNEL_RO);
- set_page_prot(empty_zero_page, PAGE_KERNEL_RO);
-
- pin_pagetable_pfn(MMUEXT_UNPIN_TABLE, PFN_DOWN(__pa(pgd)));
-
- xen_write_cr3(__pa(swapper_pg_dir));
-
- pin_pagetable_pfn(MMUEXT_PIN_L3_TABLE, PFN_DOWN(__pa(swapper_pg_dir)));
-
- return swapper_pg_dir;
-}
-#endif /* CONFIG_X86_64 */
-
/* First C function to be called on Xen boot */
asmlinkage void __init xen_start_kernel(void)
{
diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c
--- a/arch/x86/xen/mmu.c
+++ b/arch/x86/xen/mmu.c
@@ -47,6 +47,7 @@
#include <asm/tlbflush.h>
#include <asm/fixmap.h>
#include <asm/mmu_context.h>
+#include <asm/setup.h>
#include <asm/paravirt.h>
#include <asm/linkage.h>
@@ -55,6 +56,8 @@
#include <xen/page.h>
#include <xen/interface/xen.h>
+#include <xen/interface/version.h>
+#include <xen/hvc-console.h>
#include "multicalls.h"
#include "mmu.h"
@@ -114,6 +117,37 @@
#endif /* CONFIG_XEN_DEBUG_FS */
+
+/*
+ * Identity map, in addition to plain kernel map. This needs to be
+ * large enough to allocate page table pages to allocate the rest.
+ * Each page can map 2MB.
+ */
+static pte_t level1_ident_pgt[PTRS_PER_PTE * 4] __page_aligned_bss;
+
+#ifdef CONFIG_X86_64
+/* l3 pud for userspace vsyscall mapping */
+static pud_t level3_user_vsyscall[PTRS_PER_PUD] __page_aligned_bss;
+#endif /* CONFIG_X86_64 */
+
+/*
+ * Note about cr3 (pagetable base) values:
+ *
+ * xen_cr3 contains the current logical cr3 value; it contains the
+ * last set cr3. This may not be the current effective cr3, because
+ * its update may be being lazily deferred. However, a vcpu looking
+ * at its own cr3 can use this value knowing that it everything will
+ * be self-consistent.
+ *
+ * xen_current_cr3 contains the actual vcpu cr3; it is set once the
+ * hypercall to set the vcpu cr3 is complete (so it may be a little
+ * out of date, but it will never be set early). If one vcpu is
+ * looking at another vcpu''s cr3 value, it should use this variable.
+ */
+DEFINE_PER_CPU(unsigned long, xen_cr3); /* cr3 stored as physaddr */
+DEFINE_PER_CPU(unsigned long, xen_current_cr3); /* actual vcpu cr3 */
+
+
/*
* Just beyond the highest usermode address. STACK_TOP_MAX has a
* redzone above it, so round it up to a PGD boundary.
@@ -1152,6 +1186,672 @@
spin_unlock(&mm->page_table_lock);
}
+static __init void xen_pagetable_setup_start(pgd_t *base)
+{
+}
+
+static __init void xen_pagetable_setup_done(pgd_t *base)
+{
+ xen_setup_shared_info();
+}
+
+static void xen_write_cr2(unsigned long cr2)
+{
+ percpu_read(xen_vcpu)->arch.cr2 = cr2;
+}
+
+static unsigned long xen_read_cr2(void)
+{
+ return percpu_read(xen_vcpu)->arch.cr2;
+}
+
+unsigned long xen_read_cr2_direct(void)
+{
+ return percpu_read(xen_vcpu_info.arch.cr2);
+}
+
+static void xen_flush_tlb(void)
+{
+ struct mmuext_op *op;
+ struct multicall_space mcs;
+
+ preempt_disable();
+
+ mcs = xen_mc_entry(sizeof(*op));
+
+ op = mcs.args;
+ op->cmd = MMUEXT_TLB_FLUSH_LOCAL;
+ MULTI_mmuext_op(mcs.mc, op, 1, NULL, DOMID_SELF);
+
+ xen_mc_issue(PARAVIRT_LAZY_MMU);
+
+ preempt_enable();
+}
+
+static void xen_flush_tlb_single(unsigned long addr)
+{
+ struct mmuext_op *op;
+ struct multicall_space mcs;
+
+ preempt_disable();
+
+ mcs = xen_mc_entry(sizeof(*op));
+ op = mcs.args;
+ op->cmd = MMUEXT_INVLPG_LOCAL;
+ op->arg1.linear_addr = addr & PAGE_MASK;
+ MULTI_mmuext_op(mcs.mc, op, 1, NULL, DOMID_SELF);
+
+ xen_mc_issue(PARAVIRT_LAZY_MMU);
+
+ preempt_enable();
+}
+
+static void xen_flush_tlb_others(const struct cpumask *cpus,
+ struct mm_struct *mm, unsigned long va)
+{
+ struct {
+ struct mmuext_op op;
+ DECLARE_BITMAP(mask, NR_CPUS);
+ } *args;
+ struct multicall_space mcs;
+
+ BUG_ON(cpumask_empty(cpus));
+ BUG_ON(!mm);
+
+ mcs = xen_mc_entry(sizeof(*args));
+ args = mcs.args;
+ args->op.arg2.vcpumask = to_cpumask(args->mask);
+
+ /* Remove us, and any offline CPUS. */
+ cpumask_and(to_cpumask(args->mask), cpus, cpu_online_mask);
+ cpumask_clear_cpu(smp_processor_id(), to_cpumask(args->mask));
+ if (unlikely(cpumask_empty(to_cpumask(args->mask))))
+ goto issue;
+
+ if (va == TLB_FLUSH_ALL) {
+ args->op.cmd = MMUEXT_TLB_FLUSH_MULTI;
+ } else {
+ args->op.cmd = MMUEXT_INVLPG_MULTI;
+ args->op.arg1.linear_addr = va;
+ }
+
+ MULTI_mmuext_op(mcs.mc, &args->op, 1, NULL, DOMID_SELF);
+
+issue:
+ xen_mc_issue(PARAVIRT_LAZY_MMU);
+}
+
+static unsigned long xen_read_cr3(void)
+{
+ return percpu_read(xen_cr3);
+}
+
+static void set_current_cr3(void *v)
+{
+ percpu_write(xen_current_cr3, (unsigned long)v);
+}
+
+static void __xen_write_cr3(bool kernel, unsigned long cr3)
+{
+ struct mmuext_op *op;
+ struct multicall_space mcs;
+ unsigned long mfn;
+
+ if (cr3)
+ mfn = pfn_to_mfn(PFN_DOWN(cr3));
+ else
+ mfn = 0;
+
+ WARN_ON(mfn == 0 && kernel);
+
+ mcs = __xen_mc_entry(sizeof(*op));
+
+ op = mcs.args;
+ op->cmd = kernel ? MMUEXT_NEW_BASEPTR : MMUEXT_NEW_USER_BASEPTR;
+ op->arg1.mfn = mfn;
+
+ MULTI_mmuext_op(mcs.mc, op, 1, NULL, DOMID_SELF);
+
+ if (kernel) {
+ percpu_write(xen_cr3, cr3);
+
+ /* Update xen_current_cr3 once the batch has actually
+ been submitted. */
+ xen_mc_callback(set_current_cr3, (void *)cr3);
+ }
+}
+
+static void xen_write_cr3(unsigned long cr3)
+{
+ BUG_ON(preemptible());
+
+ xen_mc_batch(); /* disables interrupts */
+
+ /* Update while interrupts are disabled, so its atomic with
+ respect to ipis */
+ percpu_write(xen_cr3, cr3);
+
+ __xen_write_cr3(true, cr3);
+
+#ifdef CONFIG_X86_64
+ {
+ pgd_t *user_pgd = xen_get_user_pgd(__va(cr3));
+ if (user_pgd)
+ __xen_write_cr3(false, __pa(user_pgd));
+ else
+ __xen_write_cr3(false, 0);
+ }
+#endif
+
+ xen_mc_issue(PARAVIRT_LAZY_CPU); /* interrupts restored */
+}
+
+static int xen_pgd_alloc(struct mm_struct *mm)
+{
+ pgd_t *pgd = mm->pgd;
+ int ret = 0;
+
+ BUG_ON(PagePinned(virt_to_page(pgd)));
+
+#ifdef CONFIG_X86_64
+ {
+ struct page *page = virt_to_page(pgd);
+ pgd_t *user_pgd;
+
+ BUG_ON(page->private != 0);
+
+ ret = -ENOMEM;
+
+ user_pgd = (pgd_t *)__get_free_page(GFP_KERNEL | __GFP_ZERO);
+ page->private = (unsigned long)user_pgd;
+
+ if (user_pgd != NULL) {
+ user_pgd[pgd_index(VSYSCALL_START)] + __pgd(__pa(level3_user_vsyscall) |
_PAGE_TABLE);
+ ret = 0;
+ }
+
+ BUG_ON(PagePinned(virt_to_page(xen_get_user_pgd(pgd))));
+ }
+#endif
+
+ return ret;
+}
+
+static void xen_pgd_free(struct mm_struct *mm, pgd_t *pgd)
+{
+#ifdef CONFIG_X86_64
+ pgd_t *user_pgd = xen_get_user_pgd(pgd);
+
+ if (user_pgd)
+ free_page((unsigned long)user_pgd);
+#endif
+}
+
+
+/* Early in boot, while setting up the initial pagetable, assume
+ everything is pinned. */
+static __init void xen_alloc_pte_init(struct mm_struct *mm, unsigned long pfn)
+{
+#ifdef CONFIG_FLATMEM
+ BUG_ON(mem_map); /* should only be used early */
+#endif
+ make_lowmem_page_readonly(__va(PFN_PHYS(pfn)));
+}
+
+/* Early release_pte assumes that all pts are pinned, since there''s
+ only init_mm and anything attached to that is pinned. */
+static void xen_release_pte_init(unsigned long pfn)
+{
+ make_lowmem_page_readwrite(__va(PFN_PHYS(pfn)));
+}
+
+static void pin_pagetable_pfn(unsigned cmd, unsigned long pfn)
+{
+ struct mmuext_op op;
+ op.cmd = cmd;
+ op.arg1.mfn = pfn_to_mfn(pfn);
+ if (HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF))
+ BUG();
+}
+
+/* This needs to make sure the new pte page is pinned iff its being
+ attached to a pinned pagetable. */
+static void xen_alloc_ptpage(struct mm_struct *mm, unsigned long pfn, unsigned
level)
+{
+ struct page *page = pfn_to_page(pfn);
+
+ if (PagePinned(virt_to_page(mm->pgd))) {
+ SetPagePinned(page);
+
+ vm_unmap_aliases();
+ if (!PageHighMem(page)) {
+ make_lowmem_page_readonly(__va(PFN_PHYS((unsigned long)pfn)));
+ if (level == PT_PTE && USE_SPLIT_PTLOCKS)
+ pin_pagetable_pfn(MMUEXT_PIN_L1_TABLE, pfn);
+ } else {
+ /* make sure there are no stray mappings of
+ this page */
+ kmap_flush_unused();
+ }
+ }
+}
+
+static void xen_alloc_pte(struct mm_struct *mm, unsigned long pfn)
+{
+ xen_alloc_ptpage(mm, pfn, PT_PTE);
+}
+
+static void xen_alloc_pmd(struct mm_struct *mm, unsigned long pfn)
+{
+ xen_alloc_ptpage(mm, pfn, PT_PMD);
+}
+
+/* This should never happen until we''re OK to use struct page */
+static void xen_release_ptpage(unsigned long pfn, unsigned level)
+{
+ struct page *page = pfn_to_page(pfn);
+
+ if (PagePinned(page)) {
+ if (!PageHighMem(page)) {
+ if (level == PT_PTE && USE_SPLIT_PTLOCKS)
+ pin_pagetable_pfn(MMUEXT_UNPIN_TABLE, pfn);
+ make_lowmem_page_readwrite(__va(PFN_PHYS(pfn)));
+ }
+ ClearPagePinned(page);
+ }
+}
+
+static void xen_release_pte(unsigned long pfn)
+{
+ xen_release_ptpage(pfn, PT_PTE);
+}
+
+static void xen_release_pmd(unsigned long pfn)
+{
+ xen_release_ptpage(pfn, PT_PMD);
+}
+
+#if PAGETABLE_LEVELS == 4
+static void xen_alloc_pud(struct mm_struct *mm, unsigned long pfn)
+{
+ xen_alloc_ptpage(mm, pfn, PT_PUD);
+}
+
+static void xen_release_pud(unsigned long pfn)
+{
+ xen_release_ptpage(pfn, PT_PUD);
+}
+#endif
+
+void __init xen_reserve_top(void)
+{
+#ifdef CONFIG_X86_32
+ unsigned long top = HYPERVISOR_VIRT_START;
+ struct xen_platform_parameters pp;
+
+ if (HYPERVISOR_xen_version(XENVER_platform_parameters, &pp) == 0)
+ top = pp.virt_start;
+
+ reserve_top_address(-top);
+#endif /* CONFIG_X86_32 */
+}
+
+/*
+ * Like __va(), but returns address in the kernel mapping (which is
+ * all we have until the physical memory mapping has been set up.
+ */
+static void *__ka(phys_addr_t paddr)
+{
+#ifdef CONFIG_X86_64
+ return (void *)(paddr + __START_KERNEL_map);
+#else
+ return __va(paddr);
+#endif
+}
+
+/* Convert a machine address to physical address */
+static unsigned long m2p(phys_addr_t maddr)
+{
+ phys_addr_t paddr;
+
+ maddr &= PTE_PFN_MASK;
+ paddr = mfn_to_pfn(maddr >> PAGE_SHIFT) << PAGE_SHIFT;
+
+ return paddr;
+}
+
+/* Convert a machine address to kernel virtual */
+static void *m2v(phys_addr_t maddr)
+{
+ return __ka(m2p(maddr));
+}
+
+static void set_page_prot(void *addr, pgprot_t prot)
+{
+ unsigned long pfn = __pa(addr) >> PAGE_SHIFT;
+ pte_t pte = pfn_pte(pfn, prot);
+
+ if (HYPERVISOR_update_va_mapping((unsigned long)addr, pte, 0))
+ BUG();
+}
+
+static __init void xen_map_identity_early(pmd_t *pmd, unsigned long max_pfn)
+{
+ unsigned pmdidx, pteidx;
+ unsigned ident_pte;
+ unsigned long pfn;
+
+ ident_pte = 0;
+ pfn = 0;
+ for (pmdidx = 0; pmdidx < PTRS_PER_PMD && pfn < max_pfn;
pmdidx++) {
+ pte_t *pte_page;
+
+ /* Reuse or allocate a page of ptes */
+ if (pmd_present(pmd[pmdidx]))
+ pte_page = m2v(pmd[pmdidx].pmd);
+ else {
+ /* Check for free pte pages */
+ if (ident_pte == ARRAY_SIZE(level1_ident_pgt))
+ break;
+
+ pte_page = &level1_ident_pgt[ident_pte];
+ ident_pte += PTRS_PER_PTE;
+
+ pmd[pmdidx] = __pmd(__pa(pte_page) | _PAGE_TABLE);
+ }
+
+ /* Install mappings */
+ for (pteidx = 0; pteidx < PTRS_PER_PTE; pteidx++, pfn++) {
+ pte_t pte;
+
+ if (pfn > max_pfn_mapped)
+ max_pfn_mapped = pfn;
+
+ if (!pte_none(pte_page[pteidx]))
+ continue;
+
+ pte = pfn_pte(pfn, PAGE_KERNEL_EXEC);
+ pte_page[pteidx] = pte;
+ }
+ }
+
+ for (pteidx = 0; pteidx < ident_pte; pteidx += PTRS_PER_PTE)
+ set_page_prot(&level1_ident_pgt[pteidx], PAGE_KERNEL_RO);
+
+ set_page_prot(pmd, PAGE_KERNEL_RO);
+}
+
+#ifdef CONFIG_X86_64
+static void convert_pfn_mfn(void *v)
+{
+ pte_t *pte = v;
+ int i;
+
+ /* All levels are converted the same way, so just treat them
+ as ptes. */
+ for (i = 0; i < PTRS_PER_PTE; i++)
+ pte[i] = xen_make_pte(pte[i].pte);
+}
+
+/*
+ * Set up the inital kernel pagetable.
+ *
+ * We can construct this by grafting the Xen provided pagetable into
+ * head_64.S''s preconstructed pagetables. We copy the Xen
L2''s into
+ * level2_ident_pgt, level2_kernel_pgt and level2_fixmap_pgt. This
+ * means that only the kernel has a physical mapping to start with -
+ * but that''s enough to get __va working. We need to fill in the rest
+ * of the physical mapping once some sort of allocator has been set
+ * up.
+ */
+__init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd,
+ unsigned long max_pfn)
+{
+ pud_t *l3;
+ pmd_t *l2;
+
+ /* Zap identity mapping */
+ init_level4_pgt[0] = __pgd(0);
+
+ /* Pre-constructed entries are in pfn, so convert to mfn */
+ convert_pfn_mfn(init_level4_pgt);
+ convert_pfn_mfn(level3_ident_pgt);
+ convert_pfn_mfn(level3_kernel_pgt);
+
+ l3 = m2v(pgd[pgd_index(__START_KERNEL_map)].pgd);
+ l2 = m2v(l3[pud_index(__START_KERNEL_map)].pud);
+
+ memcpy(level2_ident_pgt, l2, sizeof(pmd_t) * PTRS_PER_PMD);
+ memcpy(level2_kernel_pgt, l2, sizeof(pmd_t) * PTRS_PER_PMD);
+
+ l3 = m2v(pgd[pgd_index(__START_KERNEL_map + PMD_SIZE)].pgd);
+ l2 = m2v(l3[pud_index(__START_KERNEL_map + PMD_SIZE)].pud);
+ memcpy(level2_fixmap_pgt, l2, sizeof(pmd_t) * PTRS_PER_PMD);
+
+ /* Set up identity map */
+ xen_map_identity_early(level2_ident_pgt, max_pfn);
+
+ /* Make pagetable pieces RO */
+ set_page_prot(init_level4_pgt, PAGE_KERNEL_RO);
+ set_page_prot(level3_ident_pgt, PAGE_KERNEL_RO);
+ set_page_prot(level3_kernel_pgt, PAGE_KERNEL_RO);
+ set_page_prot(level3_user_vsyscall, PAGE_KERNEL_RO);
+ set_page_prot(level2_kernel_pgt, PAGE_KERNEL_RO);
+ set_page_prot(level2_fixmap_pgt, PAGE_KERNEL_RO);
+
+ /* Pin down new L4 */
+ pin_pagetable_pfn(MMUEXT_PIN_L4_TABLE,
+ PFN_DOWN(__pa_symbol(init_level4_pgt)));
+
+ /* Unpin Xen-provided one */
+ pin_pagetable_pfn(MMUEXT_UNPIN_TABLE, PFN_DOWN(__pa(pgd)));
+
+ /* Switch over */
+ pgd = init_level4_pgt;
+
+ /*
+ * At this stage there can be no user pgd, and no page
+ * structure to attach it to, so make sure we just set kernel
+ * pgd.
+ */
+ xen_mc_batch();
+ __xen_write_cr3(true, __pa(pgd));
+ xen_mc_issue(PARAVIRT_LAZY_CPU);
+
+ reserve_early(__pa(xen_start_info->pt_base),
+ __pa(xen_start_info->pt_base +
+ xen_start_info->nr_pt_frames * PAGE_SIZE),
+ "XEN PAGETABLES");
+
+ return pgd;
+}
+#else /* !CONFIG_X86_64 */
+static pmd_t level2_kernel_pgt[PTRS_PER_PMD] __page_aligned_bss;
+
+__init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd,
+ unsigned long max_pfn)
+{
+ pmd_t *kernel_pmd;
+
+ init_pg_tables_start = __pa(pgd);
+ init_pg_tables_end = __pa(pgd) + xen_start_info->nr_pt_frames*PAGE_SIZE;
+ max_pfn_mapped = PFN_DOWN(init_pg_tables_end + 512*1024);
+
+ kernel_pmd = m2v(pgd[KERNEL_PGD_BOUNDARY].pgd);
+ memcpy(level2_kernel_pgt, kernel_pmd, sizeof(pmd_t) * PTRS_PER_PMD);
+
+ xen_map_identity_early(level2_kernel_pgt, max_pfn);
+
+ memcpy(swapper_pg_dir, pgd, sizeof(pgd_t) * PTRS_PER_PGD);
+ set_pgd(&swapper_pg_dir[KERNEL_PGD_BOUNDARY],
+ __pgd(__pa(level2_kernel_pgt) | _PAGE_PRESENT));
+
+ set_page_prot(level2_kernel_pgt, PAGE_KERNEL_RO);
+ set_page_prot(swapper_pg_dir, PAGE_KERNEL_RO);
+ set_page_prot(empty_zero_page, PAGE_KERNEL_RO);
+
+ pin_pagetable_pfn(MMUEXT_UNPIN_TABLE, PFN_DOWN(__pa(pgd)));
+
+ xen_write_cr3(__pa(swapper_pg_dir));
+
+ pin_pagetable_pfn(MMUEXT_PIN_L3_TABLE, PFN_DOWN(__pa(swapper_pg_dir)));
+
+ return swapper_pg_dir;
+}
+#endif /* CONFIG_X86_64 */
+
+static void xen_set_fixmap(unsigned idx, unsigned long phys, pgprot_t prot)
+{
+ pte_t pte;
+
+ phys >>= PAGE_SHIFT;
+
+ switch (idx) {
+ case FIX_BTMAP_END ... FIX_BTMAP_BEGIN:
+#ifdef CONFIG_X86_F00F_BUG
+ case FIX_F00F_IDT:
+#endif
+#ifdef CONFIG_X86_32
+ case FIX_WP_TEST:
+ case FIX_VDSO:
+# ifdef CONFIG_HIGHMEM
+ case FIX_KMAP_BEGIN ... FIX_KMAP_END:
+# endif
+#else
+ case VSYSCALL_LAST_PAGE ... VSYSCALL_FIRST_PAGE:
+#endif
+#ifdef CONFIG_X86_LOCAL_APIC
+ case FIX_APIC_BASE: /* maps dummy local APIC */
+#endif
+ pte = pfn_pte(phys, prot);
+ break;
+
+ default:
+ pte = mfn_pte(phys, prot);
+ break;
+ }
+
+ __native_set_fixmap(idx, pte);
+
+#ifdef CONFIG_X86_64
+ /* Replicate changes to map the vsyscall page into the user
+ pagetable vsyscall mapping. */
+ if (idx >= VSYSCALL_LAST_PAGE && idx <= VSYSCALL_FIRST_PAGE) {
+ unsigned long vaddr = __fix_to_virt(idx);
+ set_pte_vaddr_pud(level3_user_vsyscall, vaddr, pte);
+ }
+#endif
+}
+
+__init void xen_post_allocator_init(void)
+{
+ pv_mmu_ops.set_pte = xen_set_pte;
+ pv_mmu_ops.set_pmd = xen_set_pmd;
+ pv_mmu_ops.set_pud = xen_set_pud;
+#if PAGETABLE_LEVELS == 4
+ pv_mmu_ops.set_pgd = xen_set_pgd;
+#endif
+
+ /* This will work as long as patching hasn''t happened yet
+ (which it hasn''t) */
+ pv_mmu_ops.alloc_pte = xen_alloc_pte;
+ pv_mmu_ops.alloc_pmd = xen_alloc_pmd;
+ pv_mmu_ops.release_pte = xen_release_pte;
+ pv_mmu_ops.release_pmd = xen_release_pmd;
+#if PAGETABLE_LEVELS == 4
+ pv_mmu_ops.alloc_pud = xen_alloc_pud;
+ pv_mmu_ops.release_pud = xen_release_pud;
+#endif
+
+#ifdef CONFIG_X86_64
+ SetPagePinned(virt_to_page(level3_user_vsyscall));
+#endif
+ xen_mark_init_mm_pinned();
+}
+
+
+const struct pv_mmu_ops xen_mmu_ops __initdata = {
+ .pagetable_setup_start = xen_pagetable_setup_start,
+ .pagetable_setup_done = xen_pagetable_setup_done,
+
+ .read_cr2 = xen_read_cr2,
+ .write_cr2 = xen_write_cr2,
+
+ .read_cr3 = xen_read_cr3,
+ .write_cr3 = xen_write_cr3,
+
+ .flush_tlb_user = xen_flush_tlb,
+ .flush_tlb_kernel = xen_flush_tlb,
+ .flush_tlb_single = xen_flush_tlb_single,
+ .flush_tlb_others = xen_flush_tlb_others,
+
+ .pte_update = paravirt_nop,
+ .pte_update_defer = paravirt_nop,
+
+ .pgd_alloc = xen_pgd_alloc,
+ .pgd_free = xen_pgd_free,
+
+ .alloc_pte = xen_alloc_pte_init,
+ .release_pte = xen_release_pte_init,
+ .alloc_pmd = xen_alloc_pte_init,
+ .alloc_pmd_clone = paravirt_nop,
+ .release_pmd = xen_release_pte_init,
+
+#ifdef CONFIG_HIGHPTE
+ .kmap_atomic_pte = xen_kmap_atomic_pte,
+#endif
+
+#ifdef CONFIG_X86_64
+ .set_pte = xen_set_pte,
+#else
+ .set_pte = xen_set_pte_init,
+#endif
+ .set_pte_at = xen_set_pte_at,
+ .set_pmd = xen_set_pmd_hyper,
+
+ .ptep_modify_prot_start = __ptep_modify_prot_start,
+ .ptep_modify_prot_commit = __ptep_modify_prot_commit,
+
+ .pte_val = xen_pte_val,
+ .pgd_val = xen_pgd_val,
+
+ .make_pte = xen_make_pte,
+ .make_pgd = xen_make_pgd,
+
+#ifdef CONFIG_X86_PAE
+ .set_pte_atomic = xen_set_pte_atomic,
+ .set_pte_present = xen_set_pte_at,
+ .pte_clear = xen_pte_clear,
+ .pmd_clear = xen_pmd_clear,
+#endif /* CONFIG_X86_PAE */
+ .set_pud = xen_set_pud_hyper,
+
+ .make_pmd = xen_make_pmd,
+ .pmd_val = xen_pmd_val,
+
+#if PAGETABLE_LEVELS == 4
+ .pud_val = xen_pud_val,
+ .make_pud = xen_make_pud,
+ .set_pgd = xen_set_pgd_hyper,
+
+ .alloc_pud = xen_alloc_pte_init,
+ .release_pud = xen_release_pte_init,
+#endif /* PAGETABLE_LEVELS == 4 */
+
+ .activate_mm = xen_activate_mm,
+ .dup_mmap = xen_dup_mmap,
+ .exit_mmap = xen_exit_mmap,
+
+ .lazy_mode = {
+ .enter = paravirt_enter_lazy_mmu,
+ .leave = xen_leave_lazy,
+ },
+
+ .set_fixmap = xen_set_fixmap,
+};
+
+
#ifdef CONFIG_XEN_DEBUG_FS
static struct dentry *d_mmu_debug;
diff --git a/arch/x86/xen/mmu.h b/arch/x86/xen/mmu.h
--- a/arch/x86/xen/mmu.h
+++ b/arch/x86/xen/mmu.h
@@ -54,4 +54,7 @@
void xen_ptep_modify_prot_commit(struct mm_struct *mm, unsigned long addr,
pte_t *ptep, pte_t pte);
+unsigned long xen_read_cr2_direct(void);
+
+extern const struct pv_mmu_ops xen_mmu_ops;
#endif /* _XEN_MMU_H */
diff --git a/arch/x86/xen/xen-ops.h b/arch/x86/xen/xen-ops.h
--- a/arch/x86/xen/xen-ops.h
+++ b/arch/x86/xen/xen-ops.h
@@ -13,6 +13,7 @@
struct trap_info;
void xen_copy_trap_info(struct trap_info *traps);
+DECLARE_PER_CPU(struct vcpu_info, xen_vcpu_info);
DECLARE_PER_CPU(unsigned long, xen_cr3);
DECLARE_PER_CPU(unsigned long, xen_current_cr3);
@@ -22,6 +23,13 @@
void xen_setup_mfn_list_list(void);
void xen_setup_shared_info(void);
+void xen_setup_machphys_mapping(void);
+pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd, unsigned long max_pfn);
+void xen_ident_map_ISA(void);
+void xen_reserve_top(void);
+
+void xen_leave_lazy(void);
+void xen_post_allocator_init(void);
char * __init xen_memory_setup(void);
void __init xen_arch_setup(void);
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel
Jeremy Fitzhardinge
2009-Jan-28 22:35 UTC
[Xen-devel] [PATCH 2 of 7] x86/pvops: add a paravirt_ident functions to allow special patching
Several paravirt ops implementations simply return their arguments,
the most obvious being the make_pte/pte_val class of operations on
native.
On 32-bit, the identity function is literally a no-op, as the calling
convention uses the same registers for the first argument and return.
On 64-bit, it can be implemented with a single "mov".
This patch adds special identity functions for 32 and 64 bit argument,
and machinery to recognize them and replace them with either nops or a
mov as appropriate.
At the moment, the only users for the identity functions are the
pagetable entry conversion functions.
The result is a measureable improvement on pagetable-heavy benchmarks
(2-3%, reducing the pvops overhead from 5 to 2%).
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
---
arch/x86/include/asm/paravirt.h | 5 ++
arch/x86/kernel/paravirt.c | 75 ++++++++++++++++++++++++++++++-----
arch/x86/kernel/paravirt_patch_32.c | 12 +++++
arch/x86/kernel/paravirt_patch_64.c | 15 +++++++
4 files changed, 98 insertions(+), 9 deletions(-)
diff --git a/arch/x86/include/asm/paravirt.h b/arch/x86/include/asm/paravirt.h
--- a/arch/x86/include/asm/paravirt.h
+++ b/arch/x86/include/asm/paravirt.h
@@ -388,6 +388,8 @@
asm("start_" #ops "_" #name ": " code ";
end_" #ops "_" #name ":")
unsigned paravirt_patch_nop(void);
+unsigned paravirt_patch_ident_32(void *insnbuf, unsigned len);
+unsigned paravirt_patch_ident_64(void *insnbuf, unsigned len);
unsigned paravirt_patch_ignore(unsigned len);
unsigned paravirt_patch_call(void *insnbuf,
const void *target, u16 tgt_clobbers,
@@ -1371,6 +1373,9 @@
}
void _paravirt_nop(void);
+u32 _paravirt_ident_32(u32);
+u64 _paravirt_ident_64(u64);
+
#define paravirt_nop ((void *)_paravirt_nop)
#ifdef CONFIG_SMP
diff --git a/arch/x86/kernel/paravirt.c b/arch/x86/kernel/paravirt.c
--- a/arch/x86/kernel/paravirt.c
+++ b/arch/x86/kernel/paravirt.c
@@ -44,6 +44,17 @@
{
}
+/* identity function, which can be inlined */
+u32 _paravirt_ident_32(u32 x)
+{
+ return x;
+}
+
+u64 _paravirt_ident_64(u64 x)
+{
+ return x;
+}
+
static void __init default_banner(void)
{
printk(KERN_INFO "Booting paravirtualized kernel on %s\n",
@@ -138,9 +149,16 @@
if (opfunc == NULL)
/* If there''s no function, patch it with a ud2a (BUG) */
ret = paravirt_patch_insns(insnbuf, len, ud2a, ud2a+sizeof(ud2a));
- else if (opfunc == paravirt_nop)
+ else if (opfunc == _paravirt_nop)
/* If the operation is a nop, then nop the callsite */
ret = paravirt_patch_nop();
+
+ /* identity functions just return their single argument */
+ else if (opfunc == _paravirt_ident_32)
+ ret = paravirt_patch_ident_32(insnbuf, len);
+ else if (opfunc == _paravirt_ident_64)
+ ret = paravirt_patch_ident_64(insnbuf, len);
+
else if (type == PARAVIRT_PATCH(pv_cpu_ops.iret) ||
type == PARAVIRT_PATCH(pv_cpu_ops.irq_enable_sysexit) ||
type == PARAVIRT_PATCH(pv_cpu_ops.usergs_sysret32) ||
@@ -373,6 +391,45 @@
#endif
};
+typedef pte_t make_pte_t(pteval_t);
+typedef pmd_t make_pmd_t(pmdval_t);
+typedef pud_t make_pud_t(pudval_t);
+typedef pgd_t make_pgd_t(pgdval_t);
+
+typedef pteval_t pte_val_t(pte_t);
+typedef pmdval_t pmd_val_t(pmd_t);
+typedef pudval_t pud_val_t(pud_t);
+typedef pgdval_t pgd_val_t(pgd_t);
+
+
+#if defined(CONFIG_X86_32) && !defined(CONFIG_X86_PAE)
+/* 32-bit pagetable entries */
+#define paravirt_native_make_pte (make_pte_t *)_paravirt_ident_32
+#define paravirt_native_pte_val (pte_val_t *)_paravirt_ident_32
+
+#define paravirt_native_make_pmd (make_pmd_t *)_paravirt_ident_32
+#define paravirt_native_pmd_val (pmd_val_t *)_paravirt_ident_32
+
+#define paravirt_native_make_pud (make_pud_t *)_paravirt_ident_32
+#define paravirt_native_pud_val (pud_val_t *)_paravirt_ident_32
+
+#define paravirt_native_make_pgd (make_pgd_t *)_paravirt_ident_32
+#define paravirt_native_pgd_val (pgd_val_t *)_paravirt_ident_32
+#else
+/* 64-bit pagetable entries */
+#define paravirt_native_make_pte (make_pte_t *)_paravirt_ident_64
+#define paravirt_native_pte_val (pte_val_t *)_paravirt_ident_64
+
+#define paravirt_native_make_pmd (make_pmd_t *)_paravirt_ident_64
+#define paravirt_native_pmd_val (pmd_val_t *)_paravirt_ident_64
+
+#define paravirt_native_make_pud (make_pud_t *)_paravirt_ident_64
+#define paravirt_native_pud_val (pud_val_t *)_paravirt_ident_64
+
+#define paravirt_native_make_pgd (make_pgd_t *)_paravirt_ident_64
+#define paravirt_native_pgd_val (pgd_val_t *)_paravirt_ident_64
+#endif
+
struct pv_mmu_ops pv_mmu_ops = {
#ifndef CONFIG_X86_64
.pagetable_setup_start = native_pagetable_setup_start,
@@ -424,21 +481,21 @@
.pmd_clear = native_pmd_clear,
#endif
.set_pud = native_set_pud,
- .pmd_val = native_pmd_val,
- .make_pmd = native_make_pmd,
+ .pmd_val = paravirt_native_pmd_val,
+ .make_pmd = paravirt_native_make_pmd,
#if PAGETABLE_LEVELS == 4
- .pud_val = native_pud_val,
- .make_pud = native_make_pud,
+ .pud_val = paravirt_native_pud_val,
+ .make_pud = paravirt_native_make_pud,
.set_pgd = native_set_pgd,
#endif
#endif /* PAGETABLE_LEVELS >= 3 */
- .pte_val = native_pte_val,
- .pgd_val = native_pgd_val,
+ .pte_val = paravirt_native_pte_val,
+ .pgd_val = paravirt_native_pgd_val,
- .make_pte = native_make_pte,
- .make_pgd = native_make_pgd,
+ .make_pte = paravirt_native_make_pte,
+ .make_pgd = paravirt_native_make_pgd,
.dup_mmap = paravirt_nop,
.exit_mmap = paravirt_nop,
diff --git a/arch/x86/kernel/paravirt_patch_32.c
b/arch/x86/kernel/paravirt_patch_32.c
--- a/arch/x86/kernel/paravirt_patch_32.c
+++ b/arch/x86/kernel/paravirt_patch_32.c
@@ -12,6 +12,18 @@
DEF_NATIVE(pv_cpu_ops, clts, "clts");
DEF_NATIVE(pv_cpu_ops, read_tsc, "rdtsc");
+unsigned paravirt_patch_ident_32(void *insnbuf, unsigned len)
+{
+ /* arg in %eax, return in %eax */
+ return 0;
+}
+
+unsigned paravirt_patch_ident_64(void *insnbuf, unsigned len)
+{
+ /* arg in %edx:%eax, return in %edx:%eax */
+ return 0;
+}
+
unsigned native_patch(u8 type, u16 clobbers, void *ibuf,
unsigned long addr, unsigned len)
{
diff --git a/arch/x86/kernel/paravirt_patch_64.c
b/arch/x86/kernel/paravirt_patch_64.c
--- a/arch/x86/kernel/paravirt_patch_64.c
+++ b/arch/x86/kernel/paravirt_patch_64.c
@@ -19,6 +19,21 @@
DEF_NATIVE(pv_cpu_ops, usergs_sysret32, "swapgs; sysretl");
DEF_NATIVE(pv_cpu_ops, swapgs, "swapgs");
+DEF_NATIVE(, mov32, "mov %edi, %eax");
+DEF_NATIVE(, mov64, "mov %rdi, %rax");
+
+unsigned paravirt_patch_ident_32(void *insnbuf, unsigned len)
+{
+ return paravirt_patch_insns(insnbuf, len,
+ start__mov32, end__mov32);
+}
+
+unsigned paravirt_patch_ident_64(void *insnbuf, unsigned len)
+{
+ return paravirt_patch_insns(insnbuf, len,
+ start__mov64, end__mov64);
+}
+
unsigned native_patch(u8 type, u16 clobbers, void *ibuf,
unsigned long addr, unsigned len)
{
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel
Jeremy Fitzhardinge
2009-Jan-28 22:35 UTC
[Xen-devel] [PATCH 3 of 7] x86: fix paravirt clobber in entry_64.S
The clobber is trying to say that anything except RDI is available for clobbering, but actually clobbers everything. This hasn''t mattered because the clobbers were basically ignored, but subsequent patches will rely on them. Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com> --- arch/x86/kernel/entry_64.S | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/arch/x86/kernel/entry_64.S b/arch/x86/kernel/entry_64.S --- a/arch/x86/kernel/entry_64.S +++ b/arch/x86/kernel/entry_64.S @@ -1147,7 +1147,7 @@ CFI_STARTPROC pushf CFI_ADJUST_CFA_OFFSET 8 - DISABLE_INTERRUPTS(CLBR_ANY | ~(CLBR_RDI)) + DISABLE_INTERRUPTS(CLBR_ANY & ~CLBR_RDI) SWAPGS gs_change: movl %edi,%gs _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Jeremy Fitzhardinge
2009-Jan-28 22:35 UTC
[Xen-devel] [PATCH 4 of 7] x86/paravirt: selectively save/restore regs around pvops calls
Each asm paravirt-ops call says what registers are available for clobbering. This patch makes use of this to selectively save/restore registers around each pvops call. In many cases this significantly shrinks code size. Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com> --- arch/x86/include/asm/paravirt.h | 100 +++++++++++++++++++++++++-------------- 1 file changed, 65 insertions(+), 35 deletions(-) diff --git a/arch/x86/include/asm/paravirt.h b/arch/x86/include/asm/paravirt.h --- a/arch/x86/include/asm/paravirt.h +++ b/arch/x86/include/asm/paravirt.h @@ -12,19 +12,29 @@ #define CLBR_EAX (1 << 0) #define CLBR_ECX (1 << 1) #define CLBR_EDX (1 << 2) +#define CLBR_EDI (1 << 3) -#ifdef CONFIG_X86_64 -#define CLBR_RSI (1 << 3) -#define CLBR_RDI (1 << 4) +#ifdef CONFIG_X86_32 +/* CLBR_ANY should match all regs platform has. For i386, that''s just it */ +#define CLBR_ANY ((1 << 4) - 1) +#else +#define CLBR_RAX CLBR_EAX +#define CLBR_RCX CLBR_ECX +#define CLBR_RDX CLBR_EDX +#define CLBR_RDI CLBR_EDI +#define CLBR_RSI (1 << 4) #define CLBR_R8 (1 << 5) #define CLBR_R9 (1 << 6) #define CLBR_R10 (1 << 7) #define CLBR_R11 (1 << 8) #define CLBR_ANY ((1 << 9) - 1) + +#define CLBR_ARG_REGS (CLBR_RDI | CLBR_RSI | CLBR_RDX | \ + CLBR_RCX | CLBR_R8 | CLBR_R9) +#define CLBR_RET_REG (CLBR_RAX | CLBR_RDX) +#define CLBR_SCRATCH (CLBR_R10 | CLBR_R11) + #include <asm/desc_defs.h> -#else -/* CLBR_ANY should match all regs platform has. For i386, that''s just it */ -#define CLBR_ANY ((1 << 3) - 1) #endif /* X86_64 */ #ifndef __ASSEMBLY__ @@ -1528,33 +1538,49 @@ .popsection +#define COND_PUSH(set, mask, reg) \ + .if ((~set) & mask); push %reg; .endif +#define COND_POP(set, mask, reg) \ + .if ((~set) & mask); pop %reg; .endif + #ifdef CONFIG_X86_64 -#define PV_SAVE_REGS \ - push %rax; \ - push %rcx; \ - push %rdx; \ - push %rsi; \ - push %rdi; \ - push %r8; \ - push %r9; \ - push %r10; \ - push %r11 -#define PV_RESTORE_REGS \ - pop %r11; \ - pop %r10; \ - pop %r9; \ - pop %r8; \ - pop %rdi; \ - pop %rsi; \ - pop %rdx; \ - pop %rcx; \ - pop %rax + +#define PV_SAVE_REGS(set) \ + COND_PUSH(set, CLBR_RAX, rax); \ + COND_PUSH(set, CLBR_RCX, rcx); \ + COND_PUSH(set, CLBR_RDX, rdx); \ + COND_PUSH(set, CLBR_RSI, rsi); \ + COND_PUSH(set, CLBR_RDI, rdi); \ + COND_PUSH(set, CLBR_R8, r8); \ + COND_PUSH(set, CLBR_R9, r9); \ + COND_PUSH(set, CLBR_R10, r10); \ + COND_PUSH(set, CLBR_R11, r11) +#define PV_RESTORE_REGS(set) \ + COND_POP(set, CLBR_R11, r11); \ + COND_POP(set, CLBR_R10, r10); \ + COND_POP(set, CLBR_R9, r9); \ + COND_POP(set, CLBR_R8, r8); \ + COND_POP(set, CLBR_RDI, rdi); \ + COND_POP(set, CLBR_RSI, rsi); \ + COND_POP(set, CLBR_RDX, rdx); \ + COND_POP(set, CLBR_RCX, rcx); \ + COND_POP(set, CLBR_RAX, rax) + #define PARA_PATCH(struct, off) ((PARAVIRT_PATCH_##struct + (off)) / 8) #define PARA_SITE(ptype, clobbers, ops) _PVSITE(ptype, clobbers, ops, .quad, 8) #define PARA_INDIRECT(addr) *addr(%rip) #else -#define PV_SAVE_REGS pushl %eax; pushl %edi; pushl %ecx; pushl %edx -#define PV_RESTORE_REGS popl %edx; popl %ecx; popl %edi; popl %eax +#define PV_SAVE_REGS(set) \ + COND_PUSH(set, CLBR_EAX, eax); \ + COND_PUSH(set, CLBR_EDI, edi); \ + COND_PUSH(set, CLBR_ECX, ecx); \ + COND_PUSH(set, CLBR_EDX, edx) +#define PV_RESTORE_REGS(set) \ + COND_POP(set, CLBR_EDX, edx); \ + COND_POP(set, CLBR_ECX, ecx); \ + COND_POP(set, CLBR_EDI, edi); \ + COND_POP(set, CLBR_EAX, eax) + #define PARA_PATCH(struct, off) ((PARAVIRT_PATCH_##struct + (off)) / 4) #define PARA_SITE(ptype, clobbers, ops) _PVSITE(ptype, clobbers, ops, .long, 4) #define PARA_INDIRECT(addr) *%cs:addr @@ -1566,15 +1592,15 @@ #define DISABLE_INTERRUPTS(clobbers) \ PARA_SITE(PARA_PATCH(pv_irq_ops, PV_IRQ_irq_disable), clobbers, \ - PV_SAVE_REGS; \ + PV_SAVE_REGS(clobbers); \ call PARA_INDIRECT(pv_irq_ops+PV_IRQ_irq_disable); \ - PV_RESTORE_REGS;) \ + PV_RESTORE_REGS(clobbers);) #define ENABLE_INTERRUPTS(clobbers) \ PARA_SITE(PARA_PATCH(pv_irq_ops, PV_IRQ_irq_enable), clobbers, \ - PV_SAVE_REGS; \ + PV_SAVE_REGS(clobbers); \ call PARA_INDIRECT(pv_irq_ops+PV_IRQ_irq_enable); \ - PV_RESTORE_REGS;) + PV_RESTORE_REGS(clobbers);) #define USERGS_SYSRET32 \ PARA_SITE(PARA_PATCH(pv_cpu_ops, PV_CPU_usergs_sysret32), \ @@ -1604,11 +1630,15 @@ PARA_SITE(PARA_PATCH(pv_cpu_ops, PV_CPU_swapgs), CLBR_NONE, \ swapgs) +/* + * Note: swapgs is very special, and in practise is either going to be + * implemented with a single "swapgs" instruction or something very + * special. Either way, we don''t need to save any registers for + * it. + */ #define SWAPGS \ PARA_SITE(PARA_PATCH(pv_cpu_ops, PV_CPU_swapgs), CLBR_NONE, \ - PV_SAVE_REGS; \ - call PARA_INDIRECT(pv_cpu_ops+PV_CPU_swapgs); \ - PV_RESTORE_REGS \ + call PARA_INDIRECT(pv_cpu_ops+PV_CPU_swapgs) \ ) #define GET_CR2_INTO_RCX \ _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Jeremy Fitzhardinge
2009-Jan-28 22:35 UTC
[Xen-devel] [PATCH 5 of 7] x86/paravirt: add register-saving thunks to reduce caller register pressure
One of the problems with inserting a pile of C calls where previously
there were none is that the register pressure is greatly increased.
The C calling convention says that the caller must expect a certain
set of registers may be trashed by the callee, and that the callee can
use those registers without restriction. This includes the function
argument registers, and several others.
This patch seeks to alleviate this pressure by introducing wrapper
thunks that will do the register saving/restoring, so that the
callsite doesn''t need to worry about it, but the callee function can
be conventional compiler-generated code. In many cases (particularly
performance-sensitive cases) the callee will be in assembler anyway,
and need not use the compiler''s calling convention.
Standard calling convention is:
arguments return scratch
x86-32 eax edx ecx eax ?
x86-64 rdi rsi rdx rcx rax r8 r9 r10 r11
The thunk preserves all argument and scratch registers. The return
register is not preserved, and is available as a scratch register for
unwrapped callee code (and of course the return value).
Wrapped function pointers are themselves wrapped in a struct
paravirt_callee_save structure, in order to get some warning from the
compiler when functions with mismatched calling conventions are used.
The most common paravirt ops, both statically and dynamically, are
interrupt enable/disable/save/restore, so handle them first. This is
particularly easy since their calls are handled specially anyway.
XXX Deal with VMI. What''s their calling convention?
---
arch/x86/include/asm/paravirt.h | 126 ++++++++++++++++++++++++++++++---------
arch/x86/kernel/paravirt.c | 8 +-
arch/x86/kernel/vsmp_64.c | 12 ++-
arch/x86/lguest/boot.c | 13 ++--
arch/x86/xen/enlighten.c | 8 +-
arch/x86/xen/irq.c | 14 +++-
6 files changed, 132 insertions(+), 49 deletions(-)
diff --git a/arch/x86/include/asm/paravirt.h b/arch/x86/include/asm/paravirt.h
--- a/arch/x86/include/asm/paravirt.h
+++ b/arch/x86/include/asm/paravirt.h
@@ -17,6 +17,10 @@
#ifdef CONFIG_X86_32
/* CLBR_ANY should match all regs platform has. For i386, that''s just
it */
#define CLBR_ANY ((1 << 4) - 1)
+
+#define CLBR_ARG_REGS (CLBR_EAX | CLBR_EDX | CLBR_ECX)
+#define CLBR_RET_REG (CLBR_EAX)
+#define CLBR_SCRATCH (0)
#else
#define CLBR_RAX CLBR_EAX
#define CLBR_RCX CLBR_ECX
@@ -27,16 +31,19 @@
#define CLBR_R9 (1 << 6)
#define CLBR_R10 (1 << 7)
#define CLBR_R11 (1 << 8)
+
#define CLBR_ANY ((1 << 9) - 1)
#define CLBR_ARG_REGS (CLBR_RDI | CLBR_RSI | CLBR_RDX | \
CLBR_RCX | CLBR_R8 | CLBR_R9)
-#define CLBR_RET_REG (CLBR_RAX | CLBR_RDX)
+#define CLBR_RET_REG (CLBR_RAX)
#define CLBR_SCRATCH (CLBR_R10 | CLBR_R11)
#include <asm/desc_defs.h>
#endif /* X86_64 */
+#define CLBR_CALLEE_SAVE ((CLBR_ARG_REGS | CLBR_SCRATCH) & ~CLBR_RET_REG)
+
#ifndef __ASSEMBLY__
#include <linux/types.h>
#include <linux/cpumask.h>
@@ -50,6 +57,14 @@
struct mm_struct;
struct desc_struct;
+/*
+ * Wrapper type for pointers to code which uses the non-standard
+ * calling convention. See PV_CALL_SAVE_REGS_THUNK below.
+ */
+struct paravirt_callee_save {
+ void *func;
+};
+
/* general info */
struct pv_info {
unsigned int kernel_rpl;
@@ -199,11 +214,15 @@
* expected to use X86_EFLAGS_IF; all other bits
* returned from save_fl are undefined, and may be ignored by
* restore_fl.
+ *
+ * NOTE: These functions callers expect the callee to preserve
+ * more registers than the standard C calling convention.
*/
- unsigned long (*save_fl)(void);
- void (*restore_fl)(unsigned long);
- void (*irq_disable)(void);
- void (*irq_enable)(void);
+ struct paravirt_callee_save save_fl;
+ struct paravirt_callee_save restore_fl;
+ struct paravirt_callee_save irq_disable;
+ struct paravirt_callee_save irq_enable;
+
void (*safe_halt)(void);
void (*halt)(void);
@@ -1435,12 +1454,37 @@
__parainstructions_end[];
#ifdef CONFIG_X86_32
-#define PV_SAVE_REGS "pushl %%ecx; pushl %%edx;"
-#define PV_RESTORE_REGS "popl %%edx; popl %%ecx"
+#define PV_SAVE_REGS "pushl %ecx; pushl %edx;"
+#define PV_RESTORE_REGS "popl %edx; popl %ecx;"
+
+/* save and restore all caller-save registers, except return value */
+#define PV_SAVE_ALL_CALLER_REGS PV_SAVE_REGS
+#define PV_RESTORE_ALL_CALLER_REGS PV_RESTORE_REGS
+
#define PV_FLAGS_ARG "0"
#define PV_EXTRA_CLOBBERS
#define PV_VEXTRA_CLOBBERS
#else
+/* save and restore all caller-save registers, except return value */
+#define PV_SAVE_ALL_CALLER_REGS \
+ "push %rcx;" \
+ "push %rdx;" \
+ "push %rsi;" \
+ "push %rdi;" \
+ "push %r8;" \
+ "push %r9;" \
+ "push %r10;" \
+ "push %r11;"
+#define PV_RESTORE_ALL_CALLER_REGS \
+ "pop %r11;" \
+ "pop %r10;" \
+ "pop %r9;" \
+ "pop %r8;" \
+ "pop %rdi;" \
+ "pop %rsi;" \
+ "pop %rdx;" \
+ "pop %rcx;"
+
/* We save some registers, but all of them, that''s too much. We
clobber all
* caller saved registers but the argument parameter */
#define PV_SAVE_REGS "pushq %%rdi;"
@@ -1450,52 +1494,76 @@
#define PV_FLAGS_ARG "D"
#endif
+/*
+ * Generate a thunk around a function which saves all caller-save
+ * registers except for the return value. This allows C functions to
+ * be called from assembler code where fewer than normal registers are
+ * available. It may also help code generation around calls from C
+ * code if the common case doesn''t use many registers.
+ *
+ * When a callee is wrapped in a thunk, the caller can assume that all
+ * arg regs and all scratch registers are preserved across the
+ * call. The return value in rax/eax will not be saved, even for void
+ * functions.
+ */
+#define PV_CALLEE_SAVE_REGS_THUNK(func) \
+ extern typeof(func) __raw_callee_save_##func; \
+ static void *__##func##__ __used = func; \
+ \
+ asm(".pushsection .text;" \
+ "__raw_callee_save_" #func ": " \
+ PV_SAVE_ALL_CALLER_REGS \
+ "call " #func ";" \
+ PV_RESTORE_ALL_CALLER_REGS \
+ "ret;" \
+ ".popsection")
+
+/* Get a reference to a callee-save function */
+#define PV_CALLEE_SAVE(func) \
+ ((struct paravirt_callee_save) { __raw_callee_save_##func })
+
+/* Promise that "func" already uses the right calling convention */
+#define __PV_IS_CALLEE_SAVE(func) \
+ ((struct paravirt_callee_save) { func })
+
static inline unsigned long __raw_local_save_flags(void)
{
unsigned long f;
- asm volatile(paravirt_alt(PV_SAVE_REGS
- PARAVIRT_CALL
- PV_RESTORE_REGS)
+ asm volatile(paravirt_alt(PARAVIRT_CALL)
: "=a"(f)
: paravirt_type(pv_irq_ops.save_fl),
paravirt_clobber(CLBR_EAX)
- : "memory", "cc" PV_VEXTRA_CLOBBERS);
+ : "memory", "cc");
return f;
}
static inline void raw_local_irq_restore(unsigned long f)
{
- asm volatile(paravirt_alt(PV_SAVE_REGS
- PARAVIRT_CALL
- PV_RESTORE_REGS)
+ asm volatile(paravirt_alt(PARAVIRT_CALL)
: "=a"(f)
: PV_FLAGS_ARG(f),
paravirt_type(pv_irq_ops.restore_fl),
paravirt_clobber(CLBR_EAX)
- : "memory", "cc" PV_EXTRA_CLOBBERS);
+ : "memory", "cc");
}
static inline void raw_local_irq_disable(void)
{
- asm volatile(paravirt_alt(PV_SAVE_REGS
- PARAVIRT_CALL
- PV_RESTORE_REGS)
+ asm volatile(paravirt_alt(PARAVIRT_CALL)
:
: paravirt_type(pv_irq_ops.irq_disable),
paravirt_clobber(CLBR_EAX)
- : "memory", "eax", "cc"
PV_EXTRA_CLOBBERS);
+ : "memory", "eax", "cc");
}
static inline void raw_local_irq_enable(void)
{
- asm volatile(paravirt_alt(PV_SAVE_REGS
- PARAVIRT_CALL
- PV_RESTORE_REGS)
+ asm volatile(paravirt_alt(PARAVIRT_CALL)
:
: paravirt_type(pv_irq_ops.irq_enable),
paravirt_clobber(CLBR_EAX)
- : "memory", "eax", "cc"
PV_EXTRA_CLOBBERS);
+ : "memory", "eax", "cc");
}
static inline unsigned long __raw_local_irq_save(void)
@@ -1539,9 +1607,9 @@
#define COND_PUSH(set, mask, reg) \
- .if ((~set) & mask); push %reg; .endif
+ .if ((~(set)) & mask); push %reg; .endif
#define COND_POP(set, mask, reg) \
- .if ((~set) & mask); pop %reg; .endif
+ .if ((~(set)) & mask); pop %reg; .endif
#ifdef CONFIG_X86_64
@@ -1592,15 +1660,15 @@
#define DISABLE_INTERRUPTS(clobbers) \
PARA_SITE(PARA_PATCH(pv_irq_ops, PV_IRQ_irq_disable), clobbers, \
- PV_SAVE_REGS(clobbers); \
+ PV_SAVE_REGS(clobbers | CLBR_CALLEE_SAVE); \
call PARA_INDIRECT(pv_irq_ops+PV_IRQ_irq_disable); \
- PV_RESTORE_REGS(clobbers);)
+ PV_RESTORE_REGS(clobbers | CLBR_CALLEE_SAVE);)
#define ENABLE_INTERRUPTS(clobbers) \
PARA_SITE(PARA_PATCH(pv_irq_ops, PV_IRQ_irq_enable), clobbers, \
- PV_SAVE_REGS(clobbers); \
+ PV_SAVE_REGS(clobbers | CLBR_CALLEE_SAVE); \
call PARA_INDIRECT(pv_irq_ops+PV_IRQ_irq_enable); \
- PV_RESTORE_REGS(clobbers);)
+ PV_RESTORE_REGS(clobbers | CLBR_CALLEE_SAVE);)
#define USERGS_SYSRET32 \
PARA_SITE(PARA_PATCH(pv_cpu_ops, PV_CPU_usergs_sysret32), \
diff --git a/arch/x86/kernel/paravirt.c b/arch/x86/kernel/paravirt.c
--- a/arch/x86/kernel/paravirt.c
+++ b/arch/x86/kernel/paravirt.c
@@ -310,10 +310,10 @@
struct pv_irq_ops pv_irq_ops = {
.init_IRQ = native_init_IRQ,
- .save_fl = native_save_fl,
- .restore_fl = native_restore_fl,
- .irq_disable = native_irq_disable,
- .irq_enable = native_irq_enable,
+ .save_fl = __PV_IS_CALLEE_SAVE(native_save_fl),
+ .restore_fl = __PV_IS_CALLEE_SAVE(native_restore_fl),
+ .irq_disable = __PV_IS_CALLEE_SAVE(native_irq_disable),
+ .irq_enable = __PV_IS_CALLEE_SAVE(native_irq_enable),
.safe_halt = native_safe_halt,
.halt = native_halt,
#ifdef CONFIG_X86_64
diff --git a/arch/x86/kernel/vsmp_64.c b/arch/x86/kernel/vsmp_64.c
--- a/arch/x86/kernel/vsmp_64.c
+++ b/arch/x86/kernel/vsmp_64.c
@@ -37,6 +37,7 @@
flags &= ~X86_EFLAGS_IF;
return flags;
}
+PV_CALLEE_SAVE_REGS_THUNK(vsmp_save_fl);
static void vsmp_restore_fl(unsigned long flags)
{
@@ -46,6 +47,7 @@
flags |= X86_EFLAGS_AC;
native_restore_fl(flags);
}
+PV_CALLEE_SAVE_REGS_THUNK(vsmp_restore_fl);
static void vsmp_irq_disable(void)
{
@@ -53,6 +55,7 @@
native_restore_fl((flags & ~X86_EFLAGS_IF) | X86_EFLAGS_AC);
}
+PV_CALLEE_SAVE_REGS_THUNK(vsmp_irq_disable);
static void vsmp_irq_enable(void)
{
@@ -60,6 +63,7 @@
native_restore_fl((flags | X86_EFLAGS_IF) & (~X86_EFLAGS_AC));
}
+PV_CALLEE_SAVE_REGS_THUNK(vsmp_irq_enable);
static unsigned __init_or_module vsmp_patch(u8 type, u16 clobbers, void *ibuf,
unsigned long addr, unsigned len)
@@ -90,10 +94,10 @@
cap, ctl);
if (cap & ctl & (1 << 4)) {
/* Setup irq ops and turn on vSMP IRQ fastpath handling */
- pv_irq_ops.irq_disable = vsmp_irq_disable;
- pv_irq_ops.irq_enable = vsmp_irq_enable;
- pv_irq_ops.save_fl = vsmp_save_fl;
- pv_irq_ops.restore_fl = vsmp_restore_fl;
+ pv_irq_ops.irq_disable = PV_CALLEE_SAVE(vsmp_irq_disable);
+ pv_irq_ops.irq_enable = PV_CALLEE_SAVE(vsmp_irq_enable);
+ pv_irq_ops.save_fl = PV_CALLEE_SAVE(vsmp_save_fl);
+ pv_irq_ops.restore_fl = PV_CALLEE_SAVE(vsmp_restore_fl);
pv_init_ops.patch = vsmp_patch;
ctl &= ~(1 << 4);
diff --git a/arch/x86/lguest/boot.c b/arch/x86/lguest/boot.c
--- a/arch/x86/lguest/boot.c
+++ b/arch/x86/lguest/boot.c
@@ -173,24 +173,29 @@
{
return lguest_data.irq_enabled;
}
+PV_CALLEE_SAVE_REGS_THUNK(save_fl);
/* restore_flags() just sets the flags back to the value given. */
static void restore_fl(unsigned long flags)
{
lguest_data.irq_enabled = flags;
}
+PV_CALLEE_SAVE_REGS_THUNK(restore_fl);
/* Interrupts go off... */
static void irq_disable(void)
{
lguest_data.irq_enabled = 0;
}
+PV_CALLEE_SAVE_REGS_THUNK(irq_disable);
/* Interrupts go on... */
static void irq_enable(void)
{
lguest_data.irq_enabled = X86_EFLAGS_IF;
}
+PV_CALLEE_SAVE_REGS_THUNK(irq_enable);
+
/*:*/
/*M:003 Note that we don''t check for outstanding interrupts when we
re-enable
* them (or when we unmask an interrupt). This seems to work for the moment,
@@ -984,10 +989,10 @@
/* interrupt-related operations */
pv_irq_ops.init_IRQ = lguest_init_IRQ;
- pv_irq_ops.save_fl = save_fl;
- pv_irq_ops.restore_fl = restore_fl;
- pv_irq_ops.irq_disable = irq_disable;
- pv_irq_ops.irq_enable = irq_enable;
+ pv_irq_ops.save_fl = PV_CALLEE_SAVE(save_fl);
+ pv_irq_ops.restore_fl = PV_CALLEE_SAVE(restore_fl);
+ pv_irq_ops.irq_disable = PV_CALLEE_SAVE(irq_disable);
+ pv_irq_ops.irq_enable = PV_CALLEE_SAVE(irq_enable);
pv_irq_ops.safe_halt = lguest_safe_halt;
/* init-time operations */
diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c
--- a/arch/x86/xen/enlighten.c
+++ b/arch/x86/xen/enlighten.c
@@ -676,10 +676,10 @@
if (have_vcpu_info_placement) {
printk(KERN_INFO "Xen: using vcpu_info placement\n");
- pv_irq_ops.save_fl = xen_save_fl_direct;
- pv_irq_ops.restore_fl = xen_restore_fl_direct;
- pv_irq_ops.irq_disable = xen_irq_disable_direct;
- pv_irq_ops.irq_enable = xen_irq_enable_direct;
+ pv_irq_ops.save_fl = __PV_IS_CALLEE_SAVE(xen_save_fl_direct);
+ pv_irq_ops.restore_fl = __PV_IS_CALLEE_SAVE(xen_restore_fl_direct);
+ pv_irq_ops.irq_disable = __PV_IS_CALLEE_SAVE(xen_irq_disable_direct);
+ pv_irq_ops.irq_enable = __PV_IS_CALLEE_SAVE(xen_irq_enable_direct);
pv_mmu_ops.read_cr2 = xen_read_cr2_direct;
}
}
diff --git a/arch/x86/xen/irq.c b/arch/x86/xen/irq.c
--- a/arch/x86/xen/irq.c
+++ b/arch/x86/xen/irq.c
@@ -50,6 +50,7 @@
*/
return (-flags) & X86_EFLAGS_IF;
}
+PV_CALLEE_SAVE_REGS_THUNK(xen_save_fl);
static void xen_restore_fl(unsigned long flags)
{
@@ -76,6 +77,7 @@
xen_force_evtchn_callback();
}
}
+PV_CALLEE_SAVE_REGS_THUNK(xen_restore_fl);
static void xen_irq_disable(void)
{
@@ -86,6 +88,7 @@
percpu_read(xen_vcpu)->evtchn_upcall_mask = 1;
preempt_enable_no_resched();
}
+PV_CALLEE_SAVE_REGS_THUNK(xen_irq_disable);
static void xen_irq_enable(void)
{
@@ -106,6 +109,7 @@
if (unlikely(vcpu->evtchn_upcall_pending))
xen_force_evtchn_callback();
}
+PV_CALLEE_SAVE_REGS_THUNK(xen_irq_enable);
static void xen_safe_halt(void)
{
@@ -124,10 +128,12 @@
static const struct pv_irq_ops xen_irq_ops __initdata = {
.init_IRQ = __xen_init_IRQ,
- .save_fl = xen_save_fl,
- .restore_fl = xen_restore_fl,
- .irq_disable = xen_irq_disable,
- .irq_enable = xen_irq_enable,
+
+ .save_fl = PV_CALLEE_SAVE(xen_save_fl),
+ .restore_fl = PV_CALLEE_SAVE(xen_restore_fl),
+ .irq_disable = PV_CALLEE_SAVE(xen_irq_disable),
+ .irq_enable = PV_CALLEE_SAVE(xen_irq_enable),
+
.safe_halt = xen_safe_halt,
.halt = xen_halt,
#ifdef CONFIG_X86_64
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel
Jeremy Fitzhardinge
2009-Jan-28 22:35 UTC
[Xen-devel] [PATCH 6 of 7] x86/paravirt: implement PVOP_CALL macros for callee-save functions
Functions with the callee save calling convention clobber many fewer
registers than the normal C calling convention. Implement variants of
PVOP_V?CALL* accordingly. This only bothers with functions up to 3
args, since functions with more args may as well use the normal
calling convention.
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
---
arch/x86/include/asm/paravirt.h | 134 ++++++++++++++++++++++++++++-----------
1 file changed, 99 insertions(+), 35 deletions(-)
diff --git a/arch/x86/include/asm/paravirt.h b/arch/x86/include/asm/paravirt.h
--- a/arch/x86/include/asm/paravirt.h
+++ b/arch/x86/include/asm/paravirt.h
@@ -510,25 +510,45 @@
* makes sure the incoming and outgoing types are always correct.
*/
#ifdef CONFIG_X86_32
-#define PVOP_VCALL_ARGS unsigned long __eax, __edx, __ecx
+#define PVOP_VCALL_ARGS \
+ unsigned long __eax = __eax, __edx = __edx, __ecx = __ecx
#define PVOP_CALL_ARGS PVOP_VCALL_ARGS
+
+#define PVOP_CALL_ARG1(x) "a" ((unsigned long)(x))
+#define PVOP_CALL_ARG2(x) "d" ((unsigned long)(x))
+#define PVOP_CALL_ARG3(x) "c" ((unsigned long)(x))
+
#define PVOP_VCALL_CLOBBERS "=a" (__eax), "=d" (__edx), \
"=c" (__ecx)
#define PVOP_CALL_CLOBBERS PVOP_VCALL_CLOBBERS
+
+#define PVOP_VCALLEE_CLOBBERS "=a" (__eax)
+#define PVOP_CALLEE_CLOBBERS PVOP_VCALLEE_CLOBBERS
+
#define EXTRA_CLOBBERS
#define VEXTRA_CLOBBERS
-#else
-#define PVOP_VCALL_ARGS unsigned long __edi, __esi, __edx, __ecx
+#else /* CONFIG_X86_64 */
+#define PVOP_VCALL_ARGS \
+ unsigned long __edi = __edi, __esi = __esi, \
+ __edx = __edx, __ecx = __ecx
#define PVOP_CALL_ARGS PVOP_VCALL_ARGS, __eax
+
+#define PVOP_CALL_ARG1(x) "D" ((unsigned long)(x))
+#define PVOP_CALL_ARG2(x) "S" ((unsigned long)(x))
+#define PVOP_CALL_ARG3(x) "d" ((unsigned long)(x))
+#define PVOP_CALL_ARG4(x) "c" ((unsigned long)(x))
+
#define PVOP_VCALL_CLOBBERS "=D" (__edi), \
"=S" (__esi), "=d" (__edx), \
"=c" (__ecx)
+#define PVOP_CALL_CLOBBERS PVOP_VCALL_CLOBBERS, "=a" (__eax)
-#define PVOP_CALL_CLOBBERS PVOP_VCALL_CLOBBERS, "=a" (__eax)
+#define PVOP_VCALLEE_CLOBBERS "=a" (__eax)
+#define PVOP_CALLEE_CLOBBERS PVOP_VCALLEE_CLOBBERS
#define EXTRA_CLOBBERS , "r8", "r9", "r10",
"r11"
#define VEXTRA_CLOBBERS , "rax", "r8", "r9",
"r10", "r11"
-#endif
+#endif /* CONFIG_X86_32 */
#ifdef CONFIG_PARAVIRT_DEBUG
#define PVOP_TEST_NULL(op) BUG_ON(op == NULL)
@@ -536,10 +556,11 @@
#define PVOP_TEST_NULL(op) ((void)op)
#endif
-#define __PVOP_CALL(rettype, op, pre, post, ...) \
+#define ____PVOP_CALL(rettype, op, clbr, call_clbr, extra_clbr, \
+ pre, post, ...) \
({ \
rettype __ret; \
- PVOP_CALL_ARGS; \
+ PVOP_CALL_ARGS; \
PVOP_TEST_NULL(op); \
/* This is 32-bit specific, but is okay in 64-bit */ \
/* since this condition will never hold */ \
@@ -547,70 +568,113 @@
asm volatile(pre \
paravirt_alt(PARAVIRT_CALL) \
post \
- : PVOP_CALL_CLOBBERS \
+ : call_clbr \
: paravirt_type(op), \
- paravirt_clobber(CLBR_ANY), \
+ paravirt_clobber(clbr), \
##__VA_ARGS__ \
- : "memory", "cc" EXTRA_CLOBBERS); \
+ : "memory", "cc" extra_clbr); \
__ret = (rettype)((((u64)__edx) << 32) | __eax); \
} else { \
asm volatile(pre \
paravirt_alt(PARAVIRT_CALL) \
post \
- : PVOP_CALL_CLOBBERS \
+ : call_clbr \
: paravirt_type(op), \
- paravirt_clobber(CLBR_ANY), \
+ paravirt_clobber(clbr), \
##__VA_ARGS__ \
- : "memory", "cc" EXTRA_CLOBBERS); \
+ : "memory", "cc" extra_clbr); \
__ret = (rettype)__eax; \
} \
__ret; \
})
-#define __PVOP_VCALL(op, pre, post, ...) \
+
+#define __PVOP_CALL(rettype, op, pre, post, ...) \
+ ____PVOP_CALL(rettype, op, CLBR_ANY, PVOP_CALL_CLOBBERS, \
+ EXTRA_CLOBBERS, pre, post, ##__VA_ARGS__)
+
+#define __PVOP_CALLEESAVE(rettype, op, pre, post, ...) \
+ ____PVOP_CALL(rettype, op.func, CLBR_RET_REG, \
+ PVOP_CALLEE_CLOBBERS, , \
+ pre, post, ##__VA_ARGS__)
+
+
+#define ____PVOP_VCALL(op, clbr, call_clbr, extra_clbr, pre, post, ...) \
({ \
PVOP_VCALL_ARGS; \
PVOP_TEST_NULL(op); \
asm volatile(pre \
paravirt_alt(PARAVIRT_CALL) \
post \
- : PVOP_VCALL_CLOBBERS \
+ : call_clbr \
: paravirt_type(op), \
- paravirt_clobber(CLBR_ANY), \
+ paravirt_clobber(clbr), \
##__VA_ARGS__ \
- : "memory", "cc" VEXTRA_CLOBBERS); \
+ : "memory", "cc" extra_clbr); \
})
+#define __PVOP_VCALL(op, pre, post, ...) \
+ ____PVOP_VCALL(op, CLBR_ANY, PVOP_VCALL_CLOBBERS, \
+ VEXTRA_CLOBBERS, \
+ pre, post, ##__VA_ARGS__)
+
+#define __PVOP_VCALLEESAVE(rettype, op, pre, post, ...) \
+ ____PVOP_CALL(rettype, op.func, CLBR_RET_REG, \
+ PVOP_VCALLEE_CLOBBERS, , \
+ pre, post, ##__VA_ARGS__)
+
+
+
#define PVOP_CALL0(rettype, op) \
__PVOP_CALL(rettype, op, "", "")
#define PVOP_VCALL0(op) \
__PVOP_VCALL(op, "", "")
+#define PVOP_CALLEE0(rettype, op) \
+ __PVOP_CALLEESAVE(rettype, op, "", "")
+#define PVOP_VCALLEE0(op) \
+ __PVOP_VCALLEESAVE(op, "", "")
+
+
#define PVOP_CALL1(rettype, op, arg1) \
- __PVOP_CALL(rettype, op, "", "", "0" ((unsigned
long)(arg1)))
+ __PVOP_CALL(rettype, op, "", "", PVOP_CALL_ARG1(arg1))
#define PVOP_VCALL1(op, arg1) \
- __PVOP_VCALL(op, "", "", "0" ((unsigned
long)(arg1)))
+ __PVOP_VCALL(op, "", "", PVOP_CALL_ARG1(arg1))
+
+#define PVOP_CALLEE1(rettype, op, arg1) \
+ __PVOP_CALLEESAVE(rettype, op, "", "",
PVOP_CALL_ARG1(arg1))
+#define PVOP_VCALLEE1(op, arg1) \
+ __PVOP_VCALLEESAVE(op, "", "", PVOP_CALL_ARG1(arg1))
+
#define PVOP_CALL2(rettype, op, arg1, arg2) \
- __PVOP_CALL(rettype, op, "", "", "0" ((unsigned
long)(arg1)), \
- "1" ((unsigned long)(arg2)))
+ __PVOP_CALL(rettype, op, "", "", PVOP_CALL_ARG1(arg1), \
+ PVOP_CALL_ARG2(arg2))
#define PVOP_VCALL2(op, arg1, arg2) \
- __PVOP_VCALL(op, "", "", "0" ((unsigned
long)(arg1)), \
- "1" ((unsigned long)(arg2)))
+ __PVOP_VCALL(op, "", "", PVOP_CALL_ARG1(arg1), \
+ PVOP_CALL_ARG2(arg2))
+
+#define PVOP_CALLEE2(rettype, op, arg1, arg2) \
+ __PVOP_CALLEESAVE(rettype, op, "", "",
PVOP_CALL_ARG1(arg1), \
+ PVOP_CALL_ARG2(arg2))
+#define PVOP_VCALLEE2(op, arg1, arg2) \
+ __PVOP_VCALLEESAVE(op, "", "", PVOP_CALL_ARG1(arg1), \
+ PVOP_CALL_ARG2(arg2))
+
#define PVOP_CALL3(rettype, op, arg1, arg2, arg3) \
- __PVOP_CALL(rettype, op, "", "", "0" ((unsigned
long)(arg1)), \
- "1"((unsigned long)(arg2)), "2"((unsigned long)(arg3)))
+ __PVOP_CALL(rettype, op, "", "", PVOP_CALL_ARG1(arg1), \
+ PVOP_CALL_ARG2(arg2), PVOP_CALL_ARG3(arg3))
#define PVOP_VCALL3(op, arg1, arg2, arg3) \
- __PVOP_VCALL(op, "", "", "0" ((unsigned
long)(arg1)), \
- "1"((unsigned long)(arg2)), "2"((unsigned long)(arg3)))
+ __PVOP_VCALL(op, "", "", PVOP_CALL_ARG1(arg1), \
+ PVOP_CALL_ARG2(arg2), PVOP_CALL_ARG3(arg3))
/* This is the only difference in x86_64. We can make it much simpler */
#ifdef CONFIG_X86_32
#define PVOP_CALL4(rettype, op, arg1, arg2, arg3, arg4) \
__PVOP_CALL(rettype, op, \
"push %[_arg4];", "lea 4(%%esp),%%esp;", \
- "0" ((u32)(arg1)), "1" ((u32)(arg2)), \
- "2" ((u32)(arg3)), [_arg4] "mr" ((u32)(arg4)))
+ PVOP_CALL_ARG1(arg1), PVOP_CALL_ARG2(arg2), \
+ PVOP_CALL_ARG3(arg3), [_arg4] "mr" ((u32)(arg4)))
#define PVOP_VCALL4(op, arg1, arg2, arg3, arg4) \
__PVOP_VCALL(op, \
"push %[_arg4];", "lea 4(%%esp),%%esp;", \
@@ -618,13 +682,13 @@
"2" ((u32)(arg3)), [_arg4] "mr" ((u32)(arg4)))
#else
#define PVOP_CALL4(rettype, op, arg1, arg2, arg3, arg4) \
- __PVOP_CALL(rettype, op, "", "", "0" ((unsigned
long)(arg1)), \
- "1"((unsigned long)(arg2)), "2"((unsigned long)(arg3)), \
- "3"((unsigned long)(arg4)))
+ __PVOP_CALL(rettype, op, "", "", \
+ PVOP_CALL_ARG1(arg1), PVOP_CALL_ARG2(arg2), \
+ PVOP_CALL_ARG3(arg3), PVOP_CALL_ARG4(arg4))
#define PVOP_VCALL4(op, arg1, arg2, arg3, arg4) \
- __PVOP_VCALL(op, "", "", "0" ((unsigned
long)(arg1)), \
- "1"((unsigned long)(arg2)), "2"((unsigned long)(arg3)), \
- "3"((unsigned long)(arg4)))
+ __PVOP_VCALL(op, "", "", \
+ PVOP_CALL_ARG1(arg1), PVOP_CALL_ARG2(arg2), \
+ PVOP_CALL_ARG3(arg3), PVOP_CALL_ARG4(arg4))
#endif
static inline int paravirt_enabled(void)
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel
Jeremy Fitzhardinge
2009-Jan-28 22:35 UTC
[Xen-devel] [PATCH 7 of 7] x86/paravirt: use callee-saved convention for pte_val/make_pte/etc
In the native case, pte_val, make_pte, etc are all just identity
functions, so there''s no need to clobber a lot of registers over them.
(This changes the 32-bit callee-save calling convention to return both
EAX and EDX so functions can return 64-bit values.)
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
---
arch/x86/include/asm/paravirt.h | 78 +++++++++++++++++++--------------------
arch/x86/kernel/paravirt.c | 53 ++++++--------------------
arch/x86/xen/mmu.c | 24 ++++++++----
3 files changed, 67 insertions(+), 88 deletions(-)
diff --git a/arch/x86/include/asm/paravirt.h b/arch/x86/include/asm/paravirt.h
--- a/arch/x86/include/asm/paravirt.h
+++ b/arch/x86/include/asm/paravirt.h
@@ -19,7 +19,7 @@
#define CLBR_ANY ((1 << 4) - 1)
#define CLBR_ARG_REGS (CLBR_EAX | CLBR_EDX | CLBR_ECX)
-#define CLBR_RET_REG (CLBR_EAX)
+#define CLBR_RET_REG (CLBR_EAX | CLBR_EDX)
#define CLBR_SCRATCH (0)
#else
#define CLBR_RAX CLBR_EAX
@@ -308,11 +308,11 @@
void (*ptep_modify_prot_commit)(struct mm_struct *mm, unsigned long addr,
pte_t *ptep, pte_t pte);
- pteval_t (*pte_val)(pte_t);
- pte_t (*make_pte)(pteval_t pte);
+ struct paravirt_callee_save pte_val;
+ struct paravirt_callee_save make_pte;
- pgdval_t (*pgd_val)(pgd_t);
- pgd_t (*make_pgd)(pgdval_t pgd);
+ struct paravirt_callee_save pgd_val;
+ struct paravirt_callee_save make_pgd;
#if PAGETABLE_LEVELS >= 3
#ifdef CONFIG_X86_PAE
@@ -327,12 +327,12 @@
void (*set_pud)(pud_t *pudp, pud_t pudval);
- pmdval_t (*pmd_val)(pmd_t);
- pmd_t (*make_pmd)(pmdval_t pmd);
+ struct paravirt_callee_save pmd_val;
+ struct paravirt_callee_save make_pmd;
#if PAGETABLE_LEVELS == 4
- pudval_t (*pud_val)(pud_t);
- pud_t (*make_pud)(pudval_t pud);
+ struct paravirt_callee_save pud_val;
+ struct paravirt_callee_save make_pud;
void (*set_pgd)(pgd_t *pudp, pgd_t pgdval);
#endif /* PAGETABLE_LEVELS == 4 */
@@ -1155,13 +1155,13 @@
pteval_t ret;
if (sizeof(pteval_t) > sizeof(long))
- ret = PVOP_CALL2(pteval_t,
- pv_mmu_ops.make_pte,
- val, (u64)val >> 32);
+ ret = PVOP_CALLEE2(pteval_t,
+ pv_mmu_ops.make_pte,
+ val, (u64)val >> 32);
else
- ret = PVOP_CALL1(pteval_t,
- pv_mmu_ops.make_pte,
- val);
+ ret = PVOP_CALLEE1(pteval_t,
+ pv_mmu_ops.make_pte,
+ val);
return (pte_t) { .pte = ret };
}
@@ -1171,11 +1171,11 @@
pteval_t ret;
if (sizeof(pteval_t) > sizeof(long))
- ret = PVOP_CALL2(pteval_t, pv_mmu_ops.pte_val,
- pte.pte, (u64)pte.pte >> 32);
+ ret = PVOP_CALLEE2(pteval_t, pv_mmu_ops.pte_val,
+ pte.pte, (u64)pte.pte >> 32);
else
- ret = PVOP_CALL1(pteval_t, pv_mmu_ops.pte_val,
- pte.pte);
+ ret = PVOP_CALLEE1(pteval_t, pv_mmu_ops.pte_val,
+ pte.pte);
return ret;
}
@@ -1185,11 +1185,11 @@
pgdval_t ret;
if (sizeof(pgdval_t) > sizeof(long))
- ret = PVOP_CALL2(pgdval_t, pv_mmu_ops.make_pgd,
- val, (u64)val >> 32);
+ ret = PVOP_CALLEE2(pgdval_t, pv_mmu_ops.make_pgd,
+ val, (u64)val >> 32);
else
- ret = PVOP_CALL1(pgdval_t, pv_mmu_ops.make_pgd,
- val);
+ ret = PVOP_CALLEE1(pgdval_t, pv_mmu_ops.make_pgd,
+ val);
return (pgd_t) { ret };
}
@@ -1199,11 +1199,11 @@
pgdval_t ret;
if (sizeof(pgdval_t) > sizeof(long))
- ret = PVOP_CALL2(pgdval_t, pv_mmu_ops.pgd_val,
- pgd.pgd, (u64)pgd.pgd >> 32);
+ ret = PVOP_CALLEE2(pgdval_t, pv_mmu_ops.pgd_val,
+ pgd.pgd, (u64)pgd.pgd >> 32);
else
- ret = PVOP_CALL1(pgdval_t, pv_mmu_ops.pgd_val,
- pgd.pgd);
+ ret = PVOP_CALLEE1(pgdval_t, pv_mmu_ops.pgd_val,
+ pgd.pgd);
return ret;
}
@@ -1267,11 +1267,11 @@
pmdval_t ret;
if (sizeof(pmdval_t) > sizeof(long))
- ret = PVOP_CALL2(pmdval_t, pv_mmu_ops.make_pmd,
- val, (u64)val >> 32);
+ ret = PVOP_CALLEE2(pmdval_t, pv_mmu_ops.make_pmd,
+ val, (u64)val >> 32);
else
- ret = PVOP_CALL1(pmdval_t, pv_mmu_ops.make_pmd,
- val);
+ ret = PVOP_CALLEE1(pmdval_t, pv_mmu_ops.make_pmd,
+ val);
return (pmd_t) { ret };
}
@@ -1281,11 +1281,11 @@
pmdval_t ret;
if (sizeof(pmdval_t) > sizeof(long))
- ret = PVOP_CALL2(pmdval_t, pv_mmu_ops.pmd_val,
- pmd.pmd, (u64)pmd.pmd >> 32);
+ ret = PVOP_CALLEE2(pmdval_t, pv_mmu_ops.pmd_val,
+ pmd.pmd, (u64)pmd.pmd >> 32);
else
- ret = PVOP_CALL1(pmdval_t, pv_mmu_ops.pmd_val,
- pmd.pmd);
+ ret = PVOP_CALLEE1(pmdval_t, pv_mmu_ops.pmd_val,
+ pmd.pmd);
return ret;
}
@@ -1307,11 +1307,11 @@
pudval_t ret;
if (sizeof(pudval_t) > sizeof(long))
- ret = PVOP_CALL2(pudval_t, pv_mmu_ops.make_pud,
- val, (u64)val >> 32);
+ ret = PVOP_CALLEE2(pudval_t, pv_mmu_ops.make_pud,
+ val, (u64)val >> 32);
else
- ret = PVOP_CALL1(pudval_t, pv_mmu_ops.make_pud,
- val);
+ ret = PVOP_CALLEE1(pudval_t, pv_mmu_ops.make_pud,
+ val);
return (pud_t) { ret };
}
diff --git a/arch/x86/kernel/paravirt.c b/arch/x86/kernel/paravirt.c
--- a/arch/x86/kernel/paravirt.c
+++ b/arch/x86/kernel/paravirt.c
@@ -391,43 +391,12 @@
#endif
};
-typedef pte_t make_pte_t(pteval_t);
-typedef pmd_t make_pmd_t(pmdval_t);
-typedef pud_t make_pud_t(pudval_t);
-typedef pgd_t make_pgd_t(pgdval_t);
-
-typedef pteval_t pte_val_t(pte_t);
-typedef pmdval_t pmd_val_t(pmd_t);
-typedef pudval_t pud_val_t(pud_t);
-typedef pgdval_t pgd_val_t(pgd_t);
-
-
#if defined(CONFIG_X86_32) && !defined(CONFIG_X86_PAE)
/* 32-bit pagetable entries */
-#define paravirt_native_make_pte (make_pte_t *)_paravirt_ident_32
-#define paravirt_native_pte_val (pte_val_t *)_paravirt_ident_32
-
-#define paravirt_native_make_pmd (make_pmd_t *)_paravirt_ident_32
-#define paravirt_native_pmd_val (pmd_val_t *)_paravirt_ident_32
-
-#define paravirt_native_make_pud (make_pud_t *)_paravirt_ident_32
-#define paravirt_native_pud_val (pud_val_t *)_paravirt_ident_32
-
-#define paravirt_native_make_pgd (make_pgd_t *)_paravirt_ident_32
-#define paravirt_native_pgd_val (pgd_val_t *)_paravirt_ident_32
+#define PTE_IDENT __PV_IS_CALLEE_SAVE(_paravirt_ident_32)
#else
/* 64-bit pagetable entries */
-#define paravirt_native_make_pte (make_pte_t *)_paravirt_ident_64
-#define paravirt_native_pte_val (pte_val_t *)_paravirt_ident_64
-
-#define paravirt_native_make_pmd (make_pmd_t *)_paravirt_ident_64
-#define paravirt_native_pmd_val (pmd_val_t *)_paravirt_ident_64
-
-#define paravirt_native_make_pud (make_pud_t *)_paravirt_ident_64
-#define paravirt_native_pud_val (pud_val_t *)_paravirt_ident_64
-
-#define paravirt_native_make_pgd (make_pgd_t *)_paravirt_ident_64
-#define paravirt_native_pgd_val (pgd_val_t *)_paravirt_ident_64
+#define PTE_IDENT __PV_IS_CALLEE_SAVE(_paravirt_ident_64)
#endif
struct pv_mmu_ops pv_mmu_ops = {
@@ -481,21 +450,23 @@
.pmd_clear = native_pmd_clear,
#endif
.set_pud = native_set_pud,
- .pmd_val = paravirt_native_pmd_val,
- .make_pmd = paravirt_native_make_pmd,
+
+ .pmd_val = PTE_IDENT,
+ .make_pmd = PTE_IDENT,
#if PAGETABLE_LEVELS == 4
- .pud_val = paravirt_native_pud_val,
- .make_pud = paravirt_native_make_pud,
+ .pud_val = PTE_IDENT,
+ .make_pud = PTE_IDENT,
+
.set_pgd = native_set_pgd,
#endif
#endif /* PAGETABLE_LEVELS >= 3 */
- .pte_val = paravirt_native_pte_val,
- .pgd_val = paravirt_native_pgd_val,
+ .pte_val = PTE_IDENT,
+ .pgd_val = PTE_IDENT,
- .make_pte = paravirt_native_make_pte,
- .make_pgd = paravirt_native_make_pgd,
+ .make_pte = PTE_IDENT,
+ .make_pgd = PTE_IDENT,
.dup_mmap = paravirt_nop,
.exit_mmap = paravirt_nop,
diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c
--- a/arch/x86/xen/mmu.c
+++ b/arch/x86/xen/mmu.c
@@ -492,28 +492,33 @@
{
return pte_mfn_to_pfn(pte.pte);
}
+PV_CALLEE_SAVE_REGS_THUNK(xen_pte_val);
pgdval_t xen_pgd_val(pgd_t pgd)
{
return pte_mfn_to_pfn(pgd.pgd);
}
+PV_CALLEE_SAVE_REGS_THUNK(xen_pgd_val);
pte_t xen_make_pte(pteval_t pte)
{
pte = pte_pfn_to_mfn(pte);
return native_make_pte(pte);
}
+PV_CALLEE_SAVE_REGS_THUNK(xen_make_pte);
pgd_t xen_make_pgd(pgdval_t pgd)
{
pgd = pte_pfn_to_mfn(pgd);
return native_make_pgd(pgd);
}
+PV_CALLEE_SAVE_REGS_THUNK(xen_make_pgd);
pmdval_t xen_pmd_val(pmd_t pmd)
{
return pte_mfn_to_pfn(pmd.pmd);
}
+PV_CALLEE_SAVE_REGS_THUNK(xen_pmd_val);
void xen_set_pud_hyper(pud_t *ptr, pud_t val)
{
@@ -590,12 +595,14 @@
pmd = pte_pfn_to_mfn(pmd);
return native_make_pmd(pmd);
}
+PV_CALLEE_SAVE_REGS_THUNK(xen_make_pmd);
#if PAGETABLE_LEVELS == 4
pudval_t xen_pud_val(pud_t pud)
{
return pte_mfn_to_pfn(pud.pud);
}
+PV_CALLEE_SAVE_REGS_THUNK(xen_pud_val);
pud_t xen_make_pud(pudval_t pud)
{
@@ -603,6 +610,7 @@
return native_make_pud(pud);
}
+PV_CALLEE_SAVE_REGS_THUNK(xen_make_pud);
pgd_t *xen_get_user_pgd(pgd_t *pgd)
{
@@ -1813,11 +1821,11 @@
.ptep_modify_prot_start = __ptep_modify_prot_start,
.ptep_modify_prot_commit = __ptep_modify_prot_commit,
- .pte_val = xen_pte_val,
- .pgd_val = xen_pgd_val,
+ .pte_val = PV_CALLEE_SAVE(xen_pte_val),
+ .pgd_val = PV_CALLEE_SAVE(xen_pgd_val),
- .make_pte = xen_make_pte,
- .make_pgd = xen_make_pgd,
+ .make_pte = PV_CALLEE_SAVE(xen_make_pte),
+ .make_pgd = PV_CALLEE_SAVE(xen_make_pgd),
#ifdef CONFIG_X86_PAE
.set_pte_atomic = xen_set_pte_atomic,
@@ -1827,12 +1835,12 @@
#endif /* CONFIG_X86_PAE */
.set_pud = xen_set_pud_hyper,
- .make_pmd = xen_make_pmd,
- .pmd_val = xen_pmd_val,
+ .make_pmd = PV_CALLEE_SAVE(xen_make_pmd),
+ .pmd_val = PV_CALLEE_SAVE(xen_pmd_val),
#if PAGETABLE_LEVELS == 4
- .pud_val = xen_pud_val,
- .make_pud = xen_make_pud,
+ .pud_val = PV_CALLEE_SAVE(xen_pud_val),
+ .make_pud = PV_CALLEE_SAVE(xen_make_pud),
.set_pgd = xen_set_pgd_hyper,
.alloc_pud = xen_alloc_pte_init,
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel
Jeremy Fitzhardinge
2009-Jan-29 09:26 UTC
[Xen-devel] Re: [PATCH 2 of 7] x86/pvops: add a paravirt_ident functions to allow special patching
Rusty Russell wrote:> On Thursday 29 January 2009 09:05:02 Jeremy Fitzhardinge wrote: > >> void _paravirt_nop(void); >> +u32 _paravirt_ident_32(u32); >> +u64 _paravirt_ident_64(u64); >> + >> #define paravirt_nop ((void *)_paravirt_nop) >> > > So, we used a void * cast for the paravirt_nop case, but you decided to use explicit types for the ident cases? >Yes. Partly because any nop function is going to be basically equivalent to (void (*)(void)), but the concrete types for the ident function will vary (pointer, scalar, structure, etc).>> if (opfunc == NULL) >> /* If there''s no function, patch it with a ud2a (BUG) */ >> ret = paravirt_patch_insns(insnbuf, len, ud2a, ud2a+sizeof(ud2a)); >> - else if (opfunc == paravirt_nop) >> + else if (opfunc == _paravirt_nop) >> /* If the operation is a nop, then nop the callsite */ >> ret = paravirt_patch_nop(); >> > > Gratuitous change? >To make it consistent with the newly added lines following. And its slightly more correct.> >> +typedef pte_t make_pte_t(pteval_t); >> +typedef pmd_t make_pmd_t(pmdval_t); >> +typedef pud_t make_pud_t(pudval_t); >> +typedef pgd_t make_pgd_t(pgdval_t); >> + >> +typedef pteval_t pte_val_t(pte_t); >> +typedef pmdval_t pmd_val_t(pmd_t); >> +typedef pudval_t pud_val_t(pud_t); >> +typedef pgdval_t pgd_val_t(pgd_t); >> + >> + >> +#if defined(CONFIG_X86_32) && !defined(CONFIG_X86_PAE) >> +/* 32-bit pagetable entries */ >> +#define paravirt_native_make_pte (make_pte_t *)_paravirt_ident_32 >> +#define paravirt_native_pte_val (pte_val_t *)_paravirt_ident_32 >> + >> +#define paravirt_native_make_pmd (make_pmd_t *)_paravirt_ident_32 >> +#define paravirt_native_pmd_val (pmd_val_t *)_paravirt_ident_32 >> + >> +#define paravirt_native_make_pud (make_pud_t *)_paravirt_ident_32 >> +#define paravirt_native_pud_val (pud_val_t *)_paravirt_ident_32 >> + >> +#define paravirt_native_make_pgd (make_pgd_t *)_paravirt_ident_32 >> +#define paravirt_native_pgd_val (pgd_val_t *)_paravirt_ident_32 >> +#else >> +/* 64-bit pagetable entries */ >> +#define paravirt_native_make_pte (make_pte_t *)_paravirt_ident_64 >> +#define paravirt_native_pte_val (pte_val_t *)_paravirt_ident_64 >> + >> +#define paravirt_native_make_pmd (make_pmd_t *)_paravirt_ident_64 >> +#define paravirt_native_pmd_val (pmd_val_t *)_paravirt_ident_64 >> + >> +#define paravirt_native_make_pud (make_pud_t *)_paravirt_ident_64 >> +#define paravirt_native_pud_val (pud_val_t *)_paravirt_ident_64 >> + >> +#define paravirt_native_make_pgd (make_pgd_t *)_paravirt_ident_64 >> +#define paravirt_native_pgd_val (pgd_val_t *)_paravirt_ident_64 >> +#endif >> > > I think I prefer: > > /* make_pte etc and pgd_val etc are identity functions. */ > #define paravirt_native_page_op \ > (sizeof(pte_t) == sizeof(u64) ? paravirt_ident_64 : paravirt_ident_32) > > Then use that everywhere rather than these defines? >They disappear later in the series anyway.> But it''s a minor point; the code seems perfectly sound. >Great, thanks for reviewing. J _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Jeremy Fitzhardinge
2009-Jan-29 09:28 UTC
[Xen-devel] Re: [PATCH 3 of 7] x86: fix paravirt clobber in entry_64.S
Rusty Russell wrote:> On Thursday 29 January 2009 09:05:03 Jeremy Fitzhardinge wrote: > >> The clobber is trying to say that anything except RDI is available for >> clobbering, but actually clobbers everything. This hasn''t mattered >> because the clobbers were basically ignored, but subsequent patches >> will rely on them. >> >> Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com> >> > > Acked-by: Rusty Russell <rusty@rustcorp.com.au> > > Hmm, did we never have the deliberate-clobbering debug code for 64 bit? >No. I think I killed it off before 64-bit was even started. There was some confusion about whether the return regs were considered to be "clobbered" or not, but I don''t remember the details. J _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Jeremy Fitzhardinge
2009-Jan-29 09:30 UTC
[Xen-devel] Re: [PATCH 4 of 7] x86/paravirt: selectively save/restore regs around pvops calls
Rusty Russell wrote:> On Thursday 29 January 2009 09:05:04 Jeremy Fitzhardinge wrote: > >> +#define PV_SAVE_REGS(set) \ >> + COND_PUSH(set, CLBR_RAX, rax); \ >> + COND_PUSH(set, CLBR_RCX, rcx); \ >> + COND_PUSH(set, CLBR_RDX, rdx); \ >> + COND_PUSH(set, CLBR_RSI, rsi); \ >> + COND_PUSH(set, CLBR_RDI, rdi); \ >> + COND_PUSH(set, CLBR_R8, r8); \ >> + COND_PUSH(set, CLBR_R9, r9); \ >> + COND_PUSH(set, CLBR_R10, r10); \ >> + COND_PUSH(set, CLBR_R11, r11) >> > > It''s kind of messy. Can we do something clever here with token pasting? >I tried a few things, but no. Passing in just RAX, and then pasting CLBR_##reg works, and the assembler accepts %RAX - but - the asm code also #defines RAX <some stack offset>, so the assembler ends up seeing %48. I considered renaming all the CLBR_ defines to CLBR_rax, but that looks unhappy too. So I stayed with this. J _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Jeremy Fitzhardinge
2009-Jan-29 09:51 UTC
[Xen-devel] Re: [PATCH 0 of 7] x86/paravirt: optimise pvop calls and register use
H. Peter Anvin wrote:> The failing configuration was x86-64 "make allyesconfig". > > -hpa >Subject: x86/paravirt: fix missing callee-save call on pud_val Fix missed convertion to using callee-saved calls for pud_val, which causes a compile error when CONFIG_PARAVIRT_DEBUG is enabled. Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com> diff -r 452b0aa6f629 arch/x86/include/asm/paravirt.h --- a/arch/x86/include/asm/paravirt.h Wed Jan 28 14:18:52 2009 -0800 +++ b/arch/x86/include/asm/paravirt.h Thu Jan 29 01:49:37 2009 -0800 @@ -1321,11 +1321,11 @@ pudval_t ret; if (sizeof(pudval_t) > sizeof(long)) - ret = PVOP_CALL2(pudval_t, pv_mmu_ops.pud_val, - pud.pud, (u64)pud.pud >> 32); + ret = PVOP_CALLEE2(pudval_t, pv_mmu_ops.pud_val, + pud.pud, (u64)pud.pud >> 32); else - ret = PVOP_CALL1(pudval_t, pv_mmu_ops.pud_val, - pud.pud); + ret = PVOP_CALLEE1(pudval_t, pv_mmu_ops.pud_val, + pud.pud); return ret; } _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Jeremy Fitzhardinge
2009-Jan-29 10:46 UTC
[Xen-devel] Re: [PATCH 2 of 7] x86/pvops: add a paravirt_ident functions to allow special patching
Rusty Russell wrote:> I think I prefer: > > /* make_pte etc and pgd_val etc are identity functions. */ > #define paravirt_native_page_op \ > (sizeof(pte_t) == sizeof(u64) ? paravirt_ident_64 : paravirt_ident_32) > > Then use that everywhere rather than these defines? >Hm, also I don''t think that will work because paravirt_ident_32|64 return different types. The compiler might promote paravirt_ident_32 to returning a 64-bit number, which would be a bit disastrous. J _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Jeremy Fitzhardinge
2009-Jan-31 07:16 UTC
[Xen-devel] Re: [PATCH 0 of 7] x86/paravirt: optimise pvop calls and register use
H. Peter Anvin wrote:> Jeremy Fitzhardinge wrote: > >> H. Peter Anvin wrote: >> >>> The failing configuration was x86-64 "make allyesconfig". >>> >>> -hpa >>> >>> >> Subject: x86/paravirt: fix missing callee-save call on pud_val >> >> Fix missed convertion to using callee-saved calls for pud_val, which >> causes a compile error when CONFIG_PARAVIRT_DEBUG is enabled. >> >> > > Series applied to tip:x86/paravirt; it should make its way into > tip:master and linux-next over the next day or two, unless there are > failures.Thanks. I have a couple of additional fixes. Also, did you pick up those other two io.h patches I mailed you the other day? J _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Jeremy Fitzhardinge
2009-Jan-31 07:17 UTC
[Xen-devel] [PATCH 1/2] x86/paravirt: don''t restore second return reg
Impact: bugfix In the 32-bit calling convention, %eax:%edx is used to return 64-bit values. Don''t save and restore %edx around wrapped functions, or they can''t return a full 64-bit result. Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com> --- arch/x86/include/asm/paravirt.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) ==================================================================--- a/arch/x86/include/asm/paravirt.h +++ b/arch/x86/include/asm/paravirt.h @@ -1522,8 +1522,8 @@ #define PV_RESTORE_REGS "popl %edx; popl %ecx;" /* save and restore all caller-save registers, except return value */ -#define PV_SAVE_ALL_CALLER_REGS PV_SAVE_REGS -#define PV_RESTORE_ALL_CALLER_REGS PV_RESTORE_REGS +#define PV_SAVE_ALL_CALLER_REGS "pushl %ecx;" +#define PV_RESTORE_ALL_CALLER_REGS "popl %ecx;" #define PV_FLAGS_ARG "0" #define PV_EXTRA_CLOBBERS _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Jeremy Fitzhardinge
2009-Jan-31 07:18 UTC
[Xen-devel] [PATCH 2/2] x86/vmi: fix interrupt enable/disable/save/restore calling convention.
Zach says:> Enable/Disable have no clobbers at all. > Save clobbers only return value, %eax > Restore also clobbers nothing.This is precisely compatible with the calling convention, so we can just call them directly without wrapping. (Compile tested only.) Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com> --- arch/x86/kernel/vmi_32.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) ==================================================================--- a/arch/x86/kernel/vmi_32.c +++ b/arch/x86/kernel/vmi_32.c @@ -670,10 +670,11 @@ para_fill(pv_mmu_ops.write_cr2, SetCR2); para_fill(pv_mmu_ops.write_cr3, SetCR3); para_fill(pv_cpu_ops.write_cr4, SetCR4); - para_fill(pv_irq_ops.save_fl, GetInterruptMask); - para_fill(pv_irq_ops.restore_fl, SetInterruptMask); - para_fill(pv_irq_ops.irq_disable, DisableInterrupts); - para_fill(pv_irq_ops.irq_enable, EnableInterrupts); + + para_fill(pv_irq_ops.save_fl.func, GetInterruptMask); + para_fill(pv_irq_ops.restore_fl.func, SetInterruptMask); + para_fill(pv_irq_ops.irq_disable.func, DisableInterrupts); + para_fill(pv_irq_ops.irq_enable.func, EnableInterrupts); para_fill(pv_cpu_ops.wbinvd, WBINVD); para_fill(pv_cpu_ops.read_tsc, RDTSC); _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Ingo Molnar
2009-Jan-31 16:12 UTC
[Xen-devel] Re: [PATCH 0 of 7] x86/paravirt: optimise pvop calls and register use
* H. Peter Anvin <hpa@kernel.org> wrote:> Jeremy Fitzhardinge wrote: > > H. Peter Anvin wrote: > >> The failing configuration was x86-64 "make allyesconfig". > >> > >> -hpa > >> > > Subject: x86/paravirt: fix missing callee-save call on pud_val > > > > Fix missed convertion to using callee-saved calls for pud_val, which > > causes a compile error when CONFIG_PARAVIRT_DEBUG is enabled. > > > > Series applied to tip:x86/paravirt; it should make its way into > tip:master and linux-next over the next day or two, unless there are > failures.there''s some build trouble on VMI: arch/x86/kernel/vmi_32.c: In function ‘activate_vmi’: arch/x86/kernel/vmi_32.c:673: error: incompatible types in assignment arch/x86/kernel/vmi_32.c:673: error: incompatible types in assignment arch/x86/kernel/vmi_32.c:674: error: incompatible types in assignment arch/x86/kernel/vmi_32.c:674: error: incompatible types in assignment arch/x86/kernel/vmi_32.c:675: error: incompatible types in assignment arch/x86/kernel/vmi_32.c:675: error: incompatible types in assignment arch/x86/kernel/vmi_32.c:676: error: incompatible types in assignment arch/x86/kernel/vmi_32.c:676: error: incompatible types in assignment Ingo _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Jeremy Fitzhardinge
2009-Jan-31 17:00 UTC
[Xen-devel] Re: [PATCH 0 of 7] x86/paravirt: optimise pvop calls and register use
Ingo Molnar wrote:> there''s some build trouble on VMI: > > arch/x86/kernel/vmi_32.c: In function ‘activate_vmi’: > arch/x86/kernel/vmi_32.c:673: error: incompatible types in assignment > arch/x86/kernel/vmi_32.c:673: error: incompatible types in assignment > arch/x86/kernel/vmi_32.c:674: error: incompatible types in assignment > arch/x86/kernel/vmi_32.c:674: error: incompatible types in assignment > arch/x86/kernel/vmi_32.c:675: error: incompatible types in assignment > arch/x86/kernel/vmi_32.c:675: error: incompatible types in assignment > arch/x86/kernel/vmi_32.c:676: error: incompatible types in assignment > arch/x86/kernel/vmi_32.c:676: error: incompatible types in assignment >Yes. I sent a couple of fix patches last night, which included a fix for this. J Subject: x86/vmi: fix interrupt enable/disable/save/restore calling convention. Zach says:> Enable/Disable have no clobbers at all. > Save clobbers only return value, %eax > Restore also clobbers nothing.This is precisely compatible with the calling convention, so we can just call them directly without wrapping. Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com> --- arch/x86/kernel/vmi_32.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) ==================================================================--- a/arch/x86/kernel/vmi_32.c +++ b/arch/x86/kernel/vmi_32.c @@ -670,10 +670,11 @@ para_fill(pv_mmu_ops.write_cr2, SetCR2); para_fill(pv_mmu_ops.write_cr3, SetCR3); para_fill(pv_cpu_ops.write_cr4, SetCR4); - para_fill(pv_irq_ops.save_fl, GetInterruptMask); - para_fill(pv_irq_ops.restore_fl, SetInterruptMask); - para_fill(pv_irq_ops.irq_disable, DisableInterrupts); - para_fill(pv_irq_ops.irq_enable, EnableInterrupts); + + para_fill(pv_irq_ops.save_fl.func, GetInterruptMask); + para_fill(pv_irq_ops.restore_fl.func, SetInterruptMask); + para_fill(pv_irq_ops.irq_disable.func, DisableInterrupts); + para_fill(pv_irq_ops.irq_enable.func, EnableInterrupts); para_fill(pv_cpu_ops.wbinvd, WBINVD); para_fill(pv_cpu_ops.read_tsc, RDTSC); _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Ingo Molnar
2009-Feb-04 02:10 UTC
[Xen-devel] Re: [PATCH 0 of 7] x86/paravirt: optimise pvop calls and register use
x86/paravirt is causing boot trouble on a 32-bit box - it reboots during early bootup (it''s a laptop so no easy access to its crashlog - if there''s any). config attached. Ingo _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Jeremy Fitzhardinge
2009-Feb-04 02:16 UTC
[Xen-devel] Re: [PATCH 0 of 7] x86/paravirt: optimise pvop calls and register use
Ingo Molnar wrote:> x86/paravirt is causing boot trouble on a 32-bit box - it reboots during > early bootup (it''s a laptop so no easy access to its crashlog - if there''s > any). > > config attached.HPA just applied a patch which will likely fix this. 0eb592dbba40baebec9cdde3ff4574185de6cbcc J _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Ingo Molnar
2009-Feb-04 14:26 UTC
[Xen-devel] Re: [PATCH 0 of 7] x86/paravirt: optimise pvop calls and register use
* Jeremy Fitzhardinge <jeremy@goop.org> wrote:> Ingo Molnar wrote: >> x86/paravirt is causing boot trouble on a 32-bit box - it reboots >> during early bootup (it''s a laptop so no easy access to its crashlog - >> if there''s any). >> >> config attached. > > HPA just applied a patch which will likely fix this. > 0eb592dbba40baebec9cdde3ff4574185de6cbccthanks! Will need about 24 hours until i know for sure - the test-box that is affected by this bug triggers the problem every 10 hours or so. Ingo _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Ingo Molnar
2009-Feb-04 17:06 UTC
[Xen-devel] Re: [PATCH 0 of 7] x86/paravirt: optimise pvop calls and register use
* Ingo Molnar <mingo@elte.hu> wrote:> * Jeremy Fitzhardinge <jeremy@goop.org> wrote: > > > Ingo Molnar wrote: > >> x86/paravirt is causing boot trouble on a 32-bit box - it reboots > >> during early bootup (it''s a laptop so no easy access to its crashlog - > >> if there''s any). > >> > >> config attached. > > > > HPA just applied a patch which will likely fix this. > > 0eb592dbba40baebec9cdde3ff4574185de6cbcc > > thanks! Will need about 24 hours until i know for sure - the test-box that > is affected by this bug triggers the problem every 10 hours or so.okay - a testbox just happened to trigger this PARAVIRT=y boot crash bug and i can confirm that latest tip/master fixes it. Ingo _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Jeremy Fitzhardinge
2009-Feb-06 16:37 UTC
[Xen-devel] Re: [PATCH 5 of 7] x86/paravirt: add register-saving thunks to reduce caller register pressure
Andi Kleen wrote:> I had something like this (but not as generic) for the old pre > unification semaphores on x86-64 too. I think it''s a generally useful > facility for lots of things (in fact I''ve been asking gcc developers > for providing this in a general way for some time). >What other users do you have in mind?> So I would suggest to move the necessary macros to do that outside > paravirt.h into a separate include so that it can be used independently > and not call it PV_*.That wouldn''t be too hard to do, but I don''t think its something we should do lightly. There are quite a few pitfalls, and the bugs can get pretty obscure. It would also mean bringing out a lot of the PV_CALL machinery, and I''m not sure we really want to give that too much exposure in the light of day. J _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel