minios: general cleanup of mm.c
PATCH 3/3 minios: general cleanup of mm.c
Major clean up of mm.c. Added comments, fixed coding style, more sensible variable names, mark some local functions static etc.
---
extras/mini-os/arch/x86/mm.c +36 -28
@@ -52,20 +52,26 @@unsigned long *phys_to_machine_mapping;unsigned long mfn_zero;extern char stack[]; -extern void page_walk (unsigned long virt_addr); +extern void page_walk (unsigned long va); -void new_pt_frame (unsigned long *pt_pfn,unsigned long prev_l_mfn, -unsigned long offset,unsigned long level) +/* +* Make pt_pfn a new 'level' page table frame and hook it into the page +* table at offset in previous level MFN (pref_l_mfn). pt_pfn is a guest +* PFN. +*/ +static void new_pt_frame (unsigned long *pt_pfn,unsigned long prev_l_mfn, +unsigned long offset,unsigned long level) {pgentry_t *tab = (pgentry_t *)start_info.pt_base;unsigned long pt_page = (unsigned long )pfn_to_virt (*pt_pfn);pgentry_t prot_e, prot_t;mmu_update_t mmu_updates[1 ]; +int rc; prot_e = prot_t =0 ; -DEBUG ("Allocating new L%d pt frame for pt_pfn=%lx, " -"prev_l_mfn=%lx, offset=%lx" , - level, *pt_pfn, prev_l_mfn, offset); +DEBUG ("Allocating new L%d pt frame for pfn=%lx, " +"prev_l_mfn=%lx, offset=%lx" , + level, *pt_pfn, prev_l_mfn, offset);/* We need to clear the page, otherwise we might fail to map it as a page table page */
@@ -74,56 +80,63 @@switch ( level ) {case L1_FRAME : - prot_e =L1_PROT ; - prot_t =L2_PROT ; -break ; + prot_e =L1_PROT ; + prot_t =L2_PROT ; +break ;case L2_FRAME : - prot_e =L2_PROT ; - prot_t =L3_PROT ; -break ; + prot_e =L2_PROT ; + prot_t =L3_PROT ; +break ;#if defined(__x86_64__)case L3_FRAME : - prot_e =L3_PROT ; - prot_t =L4_PROT ; -break ; + prot_e =L3_PROT ; + prot_t =L4_PROT ; +break ;#endif default : -printk ("new_pt_frame() called with invalid level number %d\n" , level); -do_exit (); -break ; +printk ("new_pt_frame() called with invalid level number %d\n" , level); +do_exit (); +break ; } -/* Update the entry */ +/* Make PFN a page table page */ #if defined(__x86_64__) tab =pte_to_virt (tab[l4_table_offset (pt_page)]);#endif tab =pte_to_virt (tab[l3_table_offset (pt_page)]); mmu_updates[0 ].ptr = (tab[l2_table_offset (pt_page)] &PAGE_MASK ) + -sizeof (pgentry_t) *l1_table_offset (pt_page); +sizeof (pgentry_t) *l1_table_offset (pt_page); mmu_updates[0 ].val = (pgentry_t )pfn_to_mfn (*pt_pfn) <<PAGE_SHIFT | - (prot_e & ~_PAGE_RW); -if (HYPERVISOR_mmu_update (mmu_updates,1 ,NULL ,DOMID_SELF ) <0 ) + (prot_e & ~_PAGE_RW); + +if ( (rc =HYPERVISOR_mmu_update (mmu_updates,1 ,NULL ,DOMID_SELF )) <0 ) { -printk ("PTE for new page table page could not be updated\n" ); -do_exit (); +printk ("ERROR: PTE for new page table page could not be updated\n" ); +printk (" mmu_update failed with rc=%d\n" , rc); +do_exit (); } - -/* Now fill the new page table page with entries. -Update the page directory as well. */ - mmu_updates[0 ].ptr = ((pgentry_t )prev_l_mfn <<PAGE_SHIFT ) +sizeof (pgentry_t) * offset; + +/* Hook the new page table page into the hierarchy */ + mmu_updates[0 ].ptr = + ((pgentry_t )prev_l_mfn <<PAGE_SHIFT ) +sizeof (pgentry_t) * offset; mmu_updates[0 ].val = (pgentry_t )pfn_to_mfn (*pt_pfn) <<PAGE_SHIFT
| prot_t;
- if(HYPERVISOR_mmu_update(mmu_updates, 1, NULL, DOMID_SELF) < 0)
+
+ if ( (rc = HYPERVISOR_mmu_update(mmu_updates, 1, NULL, DOMID_SELF)) < 0 )
{
- printk("ERROR: mmu_update failed\n");
- do_exit();
+ printk("ERROR: mmu_update failed with rc=%d\n", rc);
+ do_exit();
}
*pt_pfn += 1;}
-/* Checks if a pagetable frame is needed (if weren't allocated by Xen) */ -static int need_pt_frame(unsigned long virt_address, int level) +/*
+ * Checks if a pagetable frame is needed at 'level' to map a given
+ * address. Note, this function is specific to the initial page table
+ * building.
+ */
+static int need_pt_frame(unsigned long va, int level)
{
unsigned long hyp_virt_start = HYPERVISOR_VIRT_START; #if defined(__x86_64__)
@@ -135,63 +148,71 @@
/* In general frames will _not_ be needed if they were already
allocated to map the hypervisor into our VA space */
#if defined(__x86_64__)
- if(level == L3_FRAME)
+ if ( level == L3_FRAME )
{
- if(l4_table_offset(virt_address) >=
- l4_table_offset(hyp_virt_start) &&
- l4_table_offset(virt_address) <=
- l4_table_offset(hyp_virt_end))
+ if ( l4_table_offset(va) >=
+ l4_table_offset(hyp_virt_start) &&
+ l4_table_offset(va) <=
+ l4_table_offset(hyp_virt_end))
return 0;
return 1;
- } else
+ }
+ else#endif
- if(level == L2_FRAME)
+ if ( level == L2_FRAME )
{
#if defined(__x86_64__)
- if(l4_table_offset(virt_address) >=
- l4_table_offset(hyp_virt_start) &&
- l4_table_offset(virt_address) <=
- l4_table_offset(hyp_virt_end))
+ if ( l4_table_offset(va) >=
+ l4_table_offset(hyp_virt_start) &&
+ l4_table_offset(va) <=
+ l4_table_offset(hyp_virt_end))
#endif
- if(l3_table_offset(virt_address) >=
- l3_table_offset(hyp_virt_start) &&
- l3_table_offset(virt_address) <=
- l3_table_offset(hyp_virt_end))
+ if ( l3_table_offset(va) >=
+ l3_table_offset(hyp_virt_start) &&
+ l3_table_offset(va) <=
+ l3_table_offset(hyp_virt_end))
return 0;
return 1;- } else
-
- /* Always need l1 frames */
- if(level == L1_FRAME)
- return 1;
+ }
+ else
+ /* Always need l1 frames */
+ if ( level == L1_FRAME )
+ return 1;
printk("ERROR: Unknown frame level %d, hypervisor %llx,%llx\n",
- level, hyp_virt_start, hyp_virt_end);
+ level, hyp_virt_start, hyp_virt_end);
return -1;}
-void build_pagetable(unsigned long *start_pfn, unsigned long *max_pfn) +/*
+ * Build the initial pagetable.
+ */
+static void build_pagetable(unsigned long *start_pfn, unsigned long *max_pfn)
{
unsigned long start_address, end_address;
unsigned long pfn_to_map, pt_pfn = *start_pfn;
static mmu_update_t mmu_updates[L1_PAGETABLE_ENTRIES + 1];
pgentry_t *tab = (pgentry_t *)start_info.pt_base, page;
- unsigned long mfn = pfn_to_mfn(virt_to_pfn(start_info.pt_base));
+ unsigned long pt_mfn = pfn_to_mfn(virt_to_pfn(start_info.pt_base));
unsigned long offset;
int count = 0;+ int rc;
- pfn_to_map = (start_info.nr_pt_frames - NOT_L1_FRAMES) * L1_PAGETABLE_ENTRIES;
+ pfn_to_map =
+ (start_info.nr_pt_frames - NOT_L1_FRAMES) * L1_PAGETABLE_ENTRIES;
- if (*max_pfn >= virt_to_pfn(HYPERVISOR_VIRT_START))
+ if ( *max_pfn >= virt_to_pfn(HYPERVISOR_VIRT_START) )
{
printk("WARNING: Mini-OS trying to use Xen virtual space. "
"Truncating memory from %dMB to ",
- ((unsigned long)pfn_to_virt(*max_pfn) - (unsigned long)&_text)>>20);
+ ((unsigned long)pfn_to_virt(*max_pfn) -
+ (unsigned long)&_text)>>20);
*max_pfn = virt_to_pfn(HYPERVISOR_VIRT_START - PAGE_SIZE);
printk("%dMB\n",
- ((unsigned long)pfn_to_virt(*max_pfn) - (unsigned long)&_text)>>20);
+ ((unsigned long)pfn_to_virt(*max_pfn) -
+ (unsigned long)&_text)>>20);
}
start_address = (unsigned long)pfn_to_virt(pfn_to_map);
@@ -200,49 +221,53 @@
/* We worked out the virtual memory range to map, now mapping loop */
printk("Mapping memory range 0x%lx - 0x%lx\n", start_address, end_address);
- while(start_address < end_address)
+ while ( start_address < end_address )
{
tab = (pgentry_t *)start_info.pt_base;- mfn = pfn_to_mfn(virt_to_pfn(start_info.pt_base));
+ pt_mfn = pfn_to_mfn(virt_to_pfn(start_info.pt_base));
#if defined(__x86_64__)
offset = l4_table_offset(start_address);
/* Need new L3 pt frame */- if(!(start_address & L3_MASK))
- if(need_pt_frame(start_address, L3_FRAME))
- new_pt_frame(&pt_pfn, mfn, offset, L3_FRAME);
+ if ( !(start_address & L3_MASK) )
+ if ( need_pt_frame(start_address, L3_FRAME) )
+ new_pt_frame(&pt_pfn, pt_mfn, offset, L3_FRAME);
page = tab[offset];
- mfn = pte_to_mfn(page);
- tab = to_virt(mfn_to_pfn(mfn) << PAGE_SHIFT);
+ pt_mfn = pte_to_mfn(page);
+ tab = to_virt(mfn_to_pfn(pt_mfn) << PAGE_SHIFT);
#endif
offset = l3_table_offset(start_address);
/* Need new L2 pt frame */- if(!(start_address & L2_MASK))
- if(need_pt_frame(start_address, L2_FRAME))
- new_pt_frame(&pt_pfn, mfn, offset, L2_FRAME);
+ if ( !(start_address & L2_MASK) )
+ if ( need_pt_frame(start_address, L2_FRAME) )
+ new_pt_frame(&pt_pfn, pt_mfn, offset, L2_FRAME);
page = tab[offset];
- mfn = pte_to_mfn(page);
- tab = to_virt(mfn_to_pfn(mfn) << PAGE_SHIFT);
+ pt_mfn = pte_to_mfn(page);
+ tab = to_virt(mfn_to_pfn(pt_mfn) << PAGE_SHIFT);
offset = l2_table_offset(start_address);
/* Need new L1 pt frame */- if(!(start_address & L1_MASK))
- if(need_pt_frame(start_address, L1_FRAME))
- new_pt_frame(&pt_pfn, mfn, offset, L1_FRAME);
+ if ( !(start_address & L1_MASK) )
+ if ( need_pt_frame(start_address, L1_FRAME) )
+ new_pt_frame(&pt_pfn, pt_mfn, offset, L1_FRAME);
page = tab[offset];
- mfn = pte_to_mfn(page);
+ pt_mfn = pte_to_mfn(page);
offset = l1_table_offset(start_address);- mmu_updates[count].ptr = ((pgentry_t)mfn << PAGE_SHIFT) + sizeof(pgentry_t) * offset;
- mmu_updates[count].val = (pgentry_t)pfn_to_mfn(pfn_to_map++) << PAGE_SHIFT | L1_PROT;
+ mmu_updates[count].ptr =
+ ((pgentry_t)pt_mfn << PAGE_SHIFT) + sizeof(pgentry_t) * offset;
+ mmu_updates[count].val =
+ (pgentry_t)pfn_to_mfn(pfn_to_map++) << PAGE_SHIFT | L1_PROT;
count++;
- if (count == L1_PAGETABLE_ENTRIES || pfn_to_map == *max_pfn)
+ if ( count == L1_PAGETABLE_ENTRIES || pfn_to_map == *max_pfn )
{
- if(HYPERVISOR_mmu_update(mmu_updates, count, NULL, DOMID_SELF) < 0)
+ rc = HYPERVISOR_mmu_update(mmu_updates, count, NULL, DOMID_SELF);
+ if ( rc < 0 )
{
- printk("PTE could not be updated\n");
+ printk("ERROR: build_pagetable(): PTE could not be updated\n");
+ printk(" mmu_update failed with rc=%d\n", rc);
do_exit();
}
count = 0;
@@ -253,20 +278,26 @@
*start_pfn = pt_pfn;}
+/*
+ * Mark portion of the address space read only.
+ */
extern void shared_info;
static void set_readonly(void *text, void *etext) {
- unsigned long start_address = ((unsigned long) text + PAGE_SIZE - 1) & PAGE_MASK;
+ unsigned long start_address =
+ ((unsigned long) text + PAGE_SIZE - 1) & PAGE_MASK;
unsigned long end_address = (unsigned long) etext;
static mmu_update_t mmu_updates[L1_PAGETABLE_ENTRIES + 1];
pgentry_t *tab = (pgentry_t *)start_info.pt_base, page;
unsigned long mfn = pfn_to_mfn(virt_to_pfn(start_info.pt_base));
unsigned long offset;
int count = 0;+ int rc;
printk("setting %p-%p readonly\n", text, etext);- while (start_address + PAGE_SIZE <= end_address) {
+ while ( start_address + PAGE_SIZE <= end_address )
+ {
tab = (pgentry_t *)start_info.pt_base;
mfn = pfn_to_mfn(virt_to_pfn(start_info.pt_base));@@ -287,20 +318,25 @@
offset = l1_table_offset(start_address);
- if (start_address != (unsigned long)&shared_info) {
- mmu_updates[count].ptr = ((pgentry_t)mfn << PAGE_SHIFT) + sizeof(pgentry_t) * offset; - mmu_updates[count].val = tab[offset] & ~_PAGE_RW; - count++;
- } else
- printk("skipped %p\n", start_address);
+ if ( start_address != (unsigned long)&shared_info )
+ {
+ mmu_updates[count].ptr =
+ ((pgentry_t)mfn << PAGE_SHIFT) + sizeof(pgentry_t) * offset;
+ mmu_updates[count].val = tab[offset] & ~_PAGE_RW;
+ count++;
+ }
+ else
+ printk("skipped %p\n", start_address);
start_address += PAGE_SIZE;
- if (count == L1_PAGETABLE_ENTRIES || start_address + PAGE_SIZE > end_address)
+ if ( count == L1_PAGETABLE_ENTRIES ||
+ start_address + PAGE_SIZE > end_address )
{
- if(HYPERVISOR_mmu_update(mmu_updates, count, NULL, DOMID_SELF) < 0)
+ rc = HYPERVISOR_mmu_update(mmu_updates, count, NULL, DOMID_SELF);
+ if ( rc < 0 )
{
- printk("PTE could not be updated\n");
+ printk("ERROR: set_readonly(): PTE could not be updated\n");
do_exit();
}
count = 0;
@@ -308,11 +344,11 @@
}
{- mmuext_op_t op = {
- .cmd = MMUEXT_TLB_FLUSH_ALL,
- };
- int count;
- HYPERVISOR_mmuext_op(&op, 1, &count, DOMID_SELF);
+ mmuext_op_t op = {
+ .cmd = MMUEXT_TLB_FLUSH_ALL,
+ };
+ int count;
+ HYPERVISOR_mmuext_op(&op, 1, &count, DOMID_SELF);
}}
@@ -371,7 +407,10 @@
}
-static pgentry_t *get_pgt(unsigned long addr)
+/*
+ * get the PTE for virtual address va if it exists. Otherwise NULL.
+ */
+static pgentry_t *get_pgt(unsigned long va)
{
unsigned long mfn;
pgentry_t *tab;
@@ -381,67 +420,78 @@
mfn = virt_to_mfn(start_info.pt_base); #if defined(__x86_64__)
- offset = l4_table_offset(addr);
- if (!(tab[offset] & _PAGE_PRESENT))
+ offset = l4_table_offset(va);
+ if ( !(tab[offset] & _PAGE_PRESENT) )
return NULL;
mfn = pte_to_mfn(tab[offset]);
tab = mfn_to_virt(mfn);
#endif
- offset = l3_table_offset(addr);
- if (!(tab[offset] & _PAGE_PRESENT))
+ offset = l3_table_offset(va);
+ if ( !(tab[offset] & _PAGE_PRESENT) )
return NULL;
mfn = pte_to_mfn(tab[offset]);
tab = mfn_to_virt(mfn);
- offset = l2_table_offset(addr);
- if (!(tab[offset] & _PAGE_PRESENT))
+ offset = l2_table_offset(va);
+ if ( !(tab[offset] & _PAGE_PRESENT) )
return NULL;
mfn = pte_to_mfn(tab[offset]);
tab = mfn_to_virt(mfn);
- offset = l1_table_offset(addr);
+ offset = l1_table_offset(va);
return &tab[offset];}
-pgentry_t *need_pgt(unsigned long addr)
+
+/*
+ * return a valid PTE for a given virtual address. If PTE does not exist,
+ * allocate page-table pages.
+ */
+pgentry_t *need_pgt(unsigned long va)
{
- unsigned long mfn;
+ unsigned long pt_mfn;
pgentry_t *tab;
unsigned long pt_pfn;
unsigned offset;
tab = (pgentry_t *)start_info.pt_base;- mfn = virt_to_mfn(start_info.pt_base);
+ pt_mfn = virt_to_mfn(start_info.pt_base);
#if defined(__x86_64__)
- offset = l4_table_offset(addr);
- if (!(tab[offset] & _PAGE_PRESENT)) {
+ offset = l4_table_offset(va);
+ if ( !(tab[offset] & _PAGE_PRESENT) )
+ {
pt_pfn = virt_to_pfn(alloc_page());
- new_pt_frame(&pt_pfn, mfn, offset, L3_FRAME);
+ new_pt_frame(&pt_pfn, pt_mfn, offset, L3_FRAME);
}
ASSERT(tab[offset] & _PAGE_PRESENT);
- mfn = pte_to_mfn(tab[offset]);
- tab = mfn_to_virt(mfn);
+ pt_mfn = pte_to_mfn(tab[offset]);
+ tab = mfn_to_virt(pt_mfn);
#endif
- offset = l3_table_offset(addr);
- if (!(tab[offset] & _PAGE_PRESENT)) {
+ offset = l3_table_offset(va);
+ if ( !(tab[offset] & _PAGE_PRESENT) )
+ {
pt_pfn = virt_to_pfn(alloc_page());
- new_pt_frame(&pt_pfn, mfn, offset, L2_FRAME);
+ new_pt_frame(&pt_pfn, pt_mfn, offset, L2_FRAME);
}
ASSERT(tab[offset] & _PAGE_PRESENT);
- mfn = pte_to_mfn(tab[offset]);
- tab = mfn_to_virt(mfn);
- offset = l2_table_offset(addr);
- if (!(tab[offset] & _PAGE_PRESENT)) {
+ pt_mfn = pte_to_mfn(tab[offset]);
+ tab = mfn_to_virt(pt_mfn);
+ offset = l2_table_offset(va);
+ if ( !(tab[offset] & _PAGE_PRESENT) )
+ {
pt_pfn = virt_to_pfn(alloc_page());
- new_pt_frame(&pt_pfn, mfn, offset, L1_FRAME);
+ new_pt_frame(&pt_pfn, pt_mfn, offset, L1_FRAME);
}
ASSERT(tab[offset] & _PAGE_PRESENT);
- mfn = pte_to_mfn(tab[offset]);
- tab = mfn_to_virt(mfn);
+ pt_mfn = pte_to_mfn(tab[offset]);
+ tab = mfn_to_virt(pt_mfn);- offset = l1_table_offset(addr);
+ offset = l1_table_offset(va);
return &tab[offset];}
+/*
+ * Reserve an area of virtual address space for mappings and Heap
+ */
static unsigned long demand_map_area_start;
#ifdef __x86_64__
#define DEMAND_MAP_PAGES ((128ULL << 30) / PAGE_SIZE)
@@ -466,7 +516,8 @@
demand_map_area_start = (unsigned long) pfn_to_virt(cur_pfn);
cur_pfn += DEMAND_MAP_PAGES;- printk("Demand map pfns at %lx-%lx.\n", demand_map_area_start, pfn_to_virt(cur_pfn));
+ printk("Demand map pfns at %lx-%lx.\n",
+ demand_map_area_start, pfn_to_virt(cur_pfn));
#ifdef HAVE_LIBC
cur_pfn++;@@ -477,93 +528,123 @@
#endif
}
+unsigned long allocate_ondemand(unsigned long n, unsigned long alignment) +{
+ unsigned long x;
+ unsigned long y = 0;
+
+ /* Find a properly aligned run of n contiguous frames */
+ for ( x = 0;
+ x <= DEMAND_MAP_PAGES - n;
+ x = (x + y + 1 + alignment - 1) & ~(alignment - 1) )
+ {
+ unsigned long addr = demand_map_area_start + x * PAGE_SIZE;
+ pgentry_t *pgt = get_pgt(addr);
+ for ( y = 0; y < n; y++, addr += PAGE_SIZE )
+ {+ if ( !(addr & L1_MASK) )
+ pgt = get_pgt(addr);
+ if ( pgt )
+ {
+ if ( *pgt & _PAGE_PRESENT )
+ break;
+ pgt++;
+ }
+ }
+ if ( y == n )
+ break;
+ }
+ if ( y != n )
+ {+ printk("Failed to find %ld frames!\n", n);
+ return 0;
+ }
+ return demand_map_area_start + x * PAGE_SIZE;
+}
+
+/*
+ * Map an array of MFNs contiguously into virtual address space starting at
+ * va. map f[i*stride]+i*increment for i in 0..n-1.
+ */
#define MAP_BATCH ((STACK_SIZE / 2) / sizeof(mmu_update_t))
-void do_map_frames(unsigned long addr,
- unsigned long *f, unsigned long n, unsigned long stride,
- unsigned long increment, domid_t id, int may_fail, unsigned long prot)
+void do_map_frames(unsigned long va,
+ unsigned long *mfns, unsigned long n,
+ unsigned long stride, unsigned long incr,
+ domid_t id, int may_fail,
+ unsigned long prot)
{
pgentry_t *pgt = NULL;
unsigned long done = 0;
unsigned long i;
int rc;- while (done < n) {
- unsigned long todo;
+ if ( !mfns )
+ {
+ printk("do_map_frames: no mfns supplied\n");
+ return;
+ }
+ DEBUG("va=%p n=0x%lx, mfns[0]=0x%lx stride=0x%lx incr=0x%lx prot=0x%lx\n",
+ va, n, mfns[0], stride, incr, prot);
+
+ while ( done < n )
+ {
+ unsigned long todo;
- if (may_fail)
- todo = 1;
- else
- todo = n - done;
+ if ( may_fail )
+ todo = 1;
+ else
+ todo = n - done;
- if (todo > MAP_BATCH)
- todo = MAP_BATCH;
+ if ( todo > MAP_BATCH )
+ todo = MAP_BATCH;
- {
- mmu_update_t mmu_updates[todo];
+ {
+ mmu_update_t mmu_updates[todo];
- for (i = 0; i < todo; i++, addr += PAGE_SIZE, pgt++) {
- if (!pgt || !(addr & L1_MASK))
- pgt = need_pgt(addr);
- mmu_updates[i].ptr = virt_to_mach(pgt);
- mmu_updates[i].val = ((pgentry_t)(f[(done + i) * stride] + (done + i) * increment) << PAGE_SHIFT) | prot;
- }
+ for ( i = 0; i < todo; i++, va += PAGE_SIZE, pgt++)
+ {
+ if ( !pgt || !(va & L1_MASK) )
+ pgt = need_pgt(va);
+
+ mmu_updates[i].ptr = virt_to_mach(pgt) | MMU_NORMAL_PT_UPDATE;
+ mmu_updates[i].val = ((pgentry_t)(mfns[(done + i) * stride] +
+ (done + i) * incr)
+ << PAGE_SHIFT) | prot;
+ }
- rc = HYPERVISOR_mmu_update(mmu_updates, todo, NULL, id);
- if (rc < 0) {
- if (may_fail)
- f[done * stride] |= 0xF0000000;
- else {
- printk("Map %ld (%lx, ...) at %p failed: %d.\n", todo, f[done * stride] + done * increment, addr, rc);
+ rc = HYPERVISOR_mmu_update(mmu_updates, todo, NULL, id);
+ if ( rc < 0 )
+ {+ if (may_fail)
+ mfns[done * stride] |= 0xF0000000;
+ else {
+ printk("Map %ld (%lx, ...) at %p failed: %d.\n",
+ todo, mfns[done * stride] + done * incr, va, rc);
do_exit();- }
- }
- }
-
- done += todo;
+ }
+ }
+ }
+ done += todo;
}}
-unsigned long allocate_ondemand(unsigned long n, unsigned long alignment) +/*
+ * Map an array of MFNs contiguous into virtual address space. Virtual
+ * addresses are allocated from the on demand area.
+ */
+void *map_frames_ex(unsigned long *mfns, unsigned long n,
+ unsigned long stride, unsigned long incr,
+ unsigned long alignment,
+ domid_t id, int may_fail, unsigned long prot)
{
- unsigned long x;
- unsigned long y = 0;
+ unsigned long va = allocate_ondemand(n, alignment);- /* Find a properly aligned run of n contiguous frames */
- for (x = 0; x <= DEMAND_MAP_PAGES - n; x = (x + y + 1 + alignment - 1) & ~(alignment - 1)) {
- unsigned long addr = demand_map_area_start + x * PAGE_SIZE;
- pgentry_t *pgt = get_pgt(addr);
- for (y = 0; y < n; y++, addr += PAGE_SIZE) {- if (!(addr & L1_MASK))
- pgt = get_pgt(addr);
- if (pgt) {
- if (*pgt & _PAGE_PRESENT)
- break;
- pgt++;
- }
- }
- if (y == n)
- break;
- }
- if (y != n) {
- printk("Failed to find %ld frames!\n", n);
- return 0;
- }
- return demand_map_area_start + x * PAGE_SIZE;
-}
-
-void *map_frames_ex(unsigned long *f, unsigned long n, unsigned long stride,
- unsigned long increment, unsigned long alignment, domid_t id,
- int may_fail, unsigned long prot)
-{
- unsigned long addr = allocate_ondemand(n, alignment);
-
- if (!addr)
+ if ( !va )
return NULL;
- /* Found it at x. Map it in. */ - do_map_frames(addr, f, n, stride, increment, id, may_fail, prot); + do_map_frames(va, mfns, n, stride, incr, id, may_fail, prot);
- return (void *)addr;
+ return (void *)va;
}
/*
@@ -632,15 +713,19 @@
}
+/*
+ * Clear some of the bootstrap memory
+ */
static void clear_bootstrap(void)
{
pte_t nullpte = { };+ int rc;
/* Use first page as the CoW zero page */
memset(&_text, 0, PAGE_SIZE);
mfn_zero = virt_to_mfn((unsigned long) &_text);- if (HYPERVISOR_update_va_mapping(0, nullpte, UVMF_INVLPG))
- printk("Unable to unmap NULL page\n");
+ if ( (rc = HYPERVISOR_update_va_mapping(0, nullpte, UVMF_INVLPG)) )
+ printk("Unable to unmap NULL page. rc=%d\n", rc);
}
void arch_init_p2m(unsigned long max_pfn)
@@ -665,19 +750,19 @@
unsigned long pfn;
l3_list = (unsigned long *)alloc_page();
- for(pfn=0; pfn<max_pfn; pfn++)
+ for ( pfn=0; pfn<max_pfn; pfn++ )
{
- if(!(pfn % (L1_P2M_ENTRIES * L2_P2M_ENTRIES)))
+ if ( !(pfn % (L1_P2M_ENTRIES * L2_P2M_ENTRIES)) )
{
l2_list = (unsigned long*)alloc_page();
- if((pfn >> L3_P2M_SHIFT) > 0)
+ if ( (pfn >> L3_P2M_SHIFT) > 0 )
{
printk("Error: Too many pfns.\n");
do_exit();
}
l3_list[(pfn >> L2_P2M_SHIFT)] = virt_to_mfn(l2_list);
}
- if(!(pfn % (L1_P2M_ENTRIES)))
+ if ( !(pfn % (L1_P2M_ENTRIES)) )
{
l1_list = (unsigned long*)alloc_page();
l2_list[(pfn >> L1_P2M_SHIFT) & L2_P2M_MASK] =@@ -696,25 +781,25 @@
unsigned long start_pfn, max_pfn, virt_pfns;
- printk(" _text: %p\n", &_text);
- printk(" _etext: %p\n", &_etext);
- printk(" _erodata: %p\n", &_erodata);
- printk(" _edata: %p\n", &_edata);
- printk(" stack start: %p\n", stack);
- printk(" _end: %p\n", &_end);
+ printk(" _text: %p(VA)\n", &_text);
+ printk(" _etext: %p(VA)\n", &_etext);
+ printk(" _erodata: %p(VA)\n", &_erodata);
+ printk(" _edata: %p(VA)\n", &_edata);
+ printk("stack start: %p(VA)\n", stack);
+ printk(" _end: %p(VA)\n", &_end); /* First page follows page table pages and 3 more pages (store
page etc) */
start_pfn = PFN_UP(to_phys(start_info.pt_base)) +
- start_info.nr_pt_frames + 3;
+ start_info.nr_pt_frames + 3;
max_pfn = start_info.nr_pages;
/* We need room for demand mapping and heap, clip available memory */
virt_pfns = DEMAND_MAP_PAGES + HEAP_PAGES;
- if (max_pfn + virt_pfns + 1 < max_pfn)
+ if ( max_pfn + virt_pfns + 1 < max_pfn )
max_pfn = -(virt_pfns + 1);- printk(" start_pfn: %lx\n", start_pfn);
- printk(" max_pfn: %lx\n", max_pfn);
+ printk(" start_pfn: %lx\n", start_pfn);
+ printk(" max_pfn: %lx\n", max_pfn); build_pagetable(&start_pfn, &max_pfn);
clear_bootstrap();extras/mini-os/arch/x86/setup.c +5 -3
@@ -63,10 +63,12 @@static shared_info_t *map_shared_info (unsigned long pa) { -if (HYPERVISOR_update_va_mapping ( - (unsigned long )shared_info,__pte (pa |7 ),UVMF_INVLPG ) ) +int rc; + +if ( (rc =HYPERVISOR_update_va_mapping ( + (unsigned long )shared_info,__pte (pa |7 ),UVMF_INVLPG )) ) { -printk ("Failed to map shared_info!!\n" ); +printk ("Failed to map shared_info!! rc=%d\n" , rc);do_exit (); }return (shared_info_t *)shared_info;
extras/mini-os/kernel.c +10 -8
@@ -490,14 +490,16 @@/* print out some useful information */ printk ("Xen Minimal OS!\n" ); -printk ("start_info: %p\n" , si); -printk (" nr_pages: %lu" , si->nr_pages); -printk (" shared_inf: %08lx\n" , si->shared_info); -printk (" pt_base: %p" , (void *)si->pt_base); -printk (" mod_start: 0x%lx\n" , si->mod_start); -printk (" mod_len: %lu\n" , si->mod_len); -printk (" flags: 0x%x\n" , (unsigned int )si->flags); -printk (" cmd_line: %s\n" , +printk (" start_info: %p(VA)\n" , si); +printk (" nr_pages: 0x%lx\n" , si->nr_pages); +printk (" shared_inf: 0x%08lx(MA)\n" , si->shared_info); +printk (" pt_base: %p(VA)\n" , (void *)si->pt_base); +printk ("nr_pt_frames: 0x%lx\n" , si->nr_pt_frames); +printk (" mfn_list: %p(VA)\n" , (void *)si->mfn_list); +printk (" mod_start: 0x%lx(VA)\n" , si->mod_start); +printk (" mod_len: %lu\n" , si->mod_len); +printk (" flags: 0x%x\n" , (unsigned int )si->flags); +printk (" cmd_line: %s\n" , si->cmd_line ? (const char *)si->cmd_line :"NULL" );/* Set up events. */
Review
Quoted 12 lines · Rolf Neugebauer, le Thu 19 Feb 2009 18:04:45 +0000, a écrit :
(quoted diff) +6
-6- printk(" _text: %p\n", &_text); - printk(" _etext: %p\n", &_etext); - printk(" _erodata: %p\n", &_erodata); - printk(" _edata: %p\n", &_edata); - printk(" stack start: %p\n", stack); - printk(" _end: %p\n", &_end); + printk(" _text: %p(VA)\n", &_text); + printk(" _etext: %p(VA)\n", &_etext); + printk(" _erodata: %p(VA)\n", &_erodata); + printk(" _edata: %p(VA)\n", &_edata); + printk("stack start: %p(VA)\n", stack); + printk(" _end: %p(VA)\n", &_end);
That's really a matter of taste :)
Samuel
PATCH 3/3 minios: general cleanup of mm.c [resend]
In reply to Rolf NeugebauerMajor clean up of mm.c. Added comments, fixed coding style, more sensible variable names, mark some local functions static etc.
---
extras/mini-os/arch/x86/mm.c +292 -207
@@ -52,20 +52,26 @@unsigned long *phys_to_machine_mapping;unsigned long mfn_zero;extern char stack[]; -extern void page_walk (unsigned long virt_addr); +extern void page_walk (unsigned long va); -void new_pt_frame (unsigned long *pt_pfn,unsigned long prev_l_mfn, -unsigned long offset,unsigned long level) +/* +* Make pt_pfn a new 'level' page table frame and hook it into the page +* table at offset in previous level MFN (pref_l_mfn). pt_pfn is a guest +* PFN. +*/ +static void new_pt_frame (unsigned long *pt_pfn,unsigned long prev_l_mfn, +unsigned long offset,unsigned long level) {pgentry_t *tab = (pgentry_t *)start_info.pt_base;unsigned long pt_page = (unsigned long )pfn_to_virt (*pt_pfn);pgentry_t prot_e, prot_t;mmu_update_t mmu_updates[1 ]; +int rc; prot_e = prot_t =0 ; -DEBUG ("Allocating new L%d pt frame for pt_pfn=%lx, " -"prev_l_mfn=%lx, offset=%lx" , - level, *pt_pfn, prev_l_mfn, offset); +DEBUG ("Allocating new L%d pt frame for pfn=%lx, " +"prev_l_mfn=%lx, offset=%lx" , + level, *pt_pfn, prev_l_mfn, offset);/* We need to clear the page, otherwise we might fail to map it as a page table page */
@@ -74,56 +80,63 @@switch ( level ) {case L1_FRAME : - prot_e =L1_PROT ; - prot_t =L2_PROT ; -break ; + prot_e =L1_PROT ; + prot_t =L2_PROT ; +break ;case L2_FRAME : - prot_e =L2_PROT ; - prot_t =L3_PROT ; -break ; + prot_e =L2_PROT ; + prot_t =L3_PROT ; +break ;#if defined(__x86_64__)case L3_FRAME : - prot_e =L3_PROT ; - prot_t =L4_PROT ; -break ; + prot_e =L3_PROT ; + prot_t =L4_PROT ; +break ;#endif default : -printk ("new_pt_frame() called with invalid level number %d\n" , level); -do_exit (); -break ; +printk ("new_pt_frame() called with invalid level number %d\n" , level); +do_exit (); +break ; } -/* Update the entry */ +/* Make PFN a page table page */ #if defined(__x86_64__) tab =pte_to_virt (tab[l4_table_offset (pt_page)]);#endif tab =pte_to_virt (tab[l3_table_offset (pt_page)]); mmu_updates[0 ].ptr = (tab[l2_table_offset (pt_page)] &PAGE_MASK ) + -sizeof (pgentry_t) *l1_table_offset (pt_page); +sizeof (pgentry_t) *l1_table_offset (pt_page); mmu_updates[0 ].val = (pgentry_t )pfn_to_mfn (*pt_pfn) <<PAGE_SHIFT | - (prot_e & ~_PAGE_RW); -if (HYPERVISOR_mmu_update (mmu_updates,1 ,NULL ,DOMID_SELF ) <0 ) + (prot_e & ~_PAGE_RW); + +if ( (rc =HYPERVISOR_mmu_update (mmu_updates,1 ,NULL ,DOMID_SELF )) <0 ) { -printk ("PTE for new page table page could not be updated\n" ); -do_exit (); +printk ("ERROR: PTE for new page table page could not be updated\n" ); +printk (" mmu_update failed with rc=%d\n" , rc); +do_exit (); } - -/* Now fill the new page table page with entries. -Update the page directory as well. */ - mmu_updates[0 ].ptr = ((pgentry_t )prev_l_mfn <<PAGE_SHIFT ) +sizeof (pgentry_t) * offset; + +/* Hook the new page table page into the hierarchy */ + mmu_updates[0 ].ptr = + ((pgentry_t )prev_l_mfn <<PAGE_SHIFT ) +sizeof (pgentry_t) * offset; mmu_updates[0 ].val = (pgentry_t )pfn_to_mfn (*pt_pfn) <<PAGE_SHIFT | prot_t; -if (HYPERVISOR_mmu_update (mmu_updates,1 ,NULL ,DOMID_SELF ) <0 ) + +if ( (rc =HYPERVISOR_mmu_update (mmu_updates,1 ,NULL ,DOMID_SELF )) <0 ) { -printk ("ERROR: mmu_update failed\n" ); -do_exit (); +printk ("ERROR: mmu_update failed with rc=%d\n" , rc); +do_exit (); } *pt_pfn +=1 ; } -/* Checks if a pagetable frame is needed (if weren't allocated by Xen) */ -static int need_pt_frame (unsigned long virt_address,int level) +/* +* Checks if a pagetable frame is needed at 'level' to map a given +* address. Note, this function is specific to the initial page table +* building. +*/ +static int need_pt_frame (unsigned long va,int level) {unsigned long hyp_virt_start =HYPERVISOR_VIRT_START ;#if defined(__x86_64__)
@@ -135,63 +148,71 @@/* In general frames will _not_ be needed if they were already allocated to map the hypervisor into our VA space */ #if defined(__x86_64__) -if (level ==L3_FRAME ) +if ( level ==L3_FRAME ) { -if (l4_table_offset (virt_address) >= -l4_table_offset (hyp_virt_start) && -l4_table_offset (virt_address) <= -l4_table_offset (hyp_virt_end)) +if (l4_table_offset (va) >= +l4_table_offset (hyp_virt_start) && +l4_table_offset (va) <= +l4_table_offset (hyp_virt_end))return 0 ;return 1 ; - } else + } + else#endif -if (level ==L2_FRAME ) +if ( level ==L2_FRAME ) {#if defined(__x86_64__) -if (l4_table_offset (virt_address) >= -l4_table_offset (hyp_virt_start) && -l4_table_offset (virt_address) <= - l4_table_offset(hyp_virt_end)) +if (l4_table_offset (va) >= +l4_table_offset (hyp_virt_start) && +l4_table_offset (va) <= + l4_table_offset(hyp_virt_end))#endif -if (l3_table_offset (virt_address) >= -l3_table_offset (hyp_virt_start) && -l3_table_offset (virt_address) <= -l3_table_offset (hyp_virt_end)) +if (l3_table_offset (va) >= +l3_table_offset (hyp_virt_start) && +l3_table_offset (va) <= +l3_table_offset (hyp_virt_end))return 0 ;return 1 ; - }else - -/* Always need l1 frames */ -if (level ==L1_FRAME ) -return 1 ; + } +else +/* Always need l1 frames */ +if ( level ==L1_FRAME ) +return 1 ;printk ("ERROR: Unknown frame level %d, hypervisor %llx,%llx\n" , - level, hyp_virt_start, hyp_virt_end); + level, hyp_virt_start, hyp_virt_end);return -1 ; } -void build_pagetable (unsigned long *start_pfn,unsigned long *max_pfn) +/* +* Build the initial pagetable. +*/ +static void build_pagetable (unsigned long *start_pfn,unsigned long *max_pfn) {unsigned long start_address, end_address;unsigned long pfn_to_map, pt_pfn = *start_pfn;static mmu_update_t mmu_updates[L1_PAGETABLE_ENTRIES +1 ];pgentry_t *tab = (pgentry_t *)start_info.pt_base, page; -unsigned long mfn =pfn_to_mfn (virt_to_pfn (start_info.pt_base)); +unsigned long pt_mfn =pfn_to_mfn (virt_to_pfn (start_info.pt_base));unsigned long offset;int count =0 ; +int rc; - pfn_to_map = (start_info.nr_pt_frames -NOT_L1_FRAMES ) *L1_PAGETABLE_ENTRIES ; + pfn_to_map = + (start_info.nr_pt_frames -NOT_L1_FRAMES ) *L1_PAGETABLE_ENTRIES ; -if (*max_pfn >=virt_to_pfn (HYPERVISOR_VIRT_START )) +if ( *max_pfn >=virt_to_pfn (HYPERVISOR_VIRT_START ) ) {printk ("WARNING: Mini-OS trying to use Xen virtual space. " "Truncating memory from %dMB to " , - ((unsigned long )pfn_to_virt (*max_pfn) - (unsigned long )&_text)>>20 ); + ((unsigned long )pfn_to_virt (*max_pfn) - + (unsigned long )&_text)>>20 ); *max_pfn =virt_to_pfn (HYPERVISOR_VIRT_START -PAGE_SIZE );printk ("%dMB\n" , - ((unsigned long )pfn_to_virt (*max_pfn) - (unsigned long )&_text)>>20 ); + ((unsigned long )pfn_to_virt (*max_pfn) - + (unsigned long )&_text)>>20 ); } start_address = (unsigned long )pfn_to_virt (pfn_to_map);
@@ -200,49 +221,53 @@/* We worked out the virtual memory range to map, now mapping loop */ printk ("Mapping memory range 0x%lx - 0x%lx\n" , start_address, end_address); -while (start_address < end_address) +while ( start_address < end_address ) { tab = (pgentry_t *)start_info.pt_base; - mfn =pfn_to_mfn (virt_to_pfn (start_info.pt_base)); + pt_mfn =pfn_to_mfn (virt_to_pfn (start_info.pt_base));#if defined(__x86_64__) offset =l4_table_offset (start_address);/* Need new L3 pt frame */ -if (!(start_address &L3_MASK )) -if (need_pt_frame (start_address,L3_FRAME )) -new_pt_frame (&pt_pfn, mfn, offset,L3_FRAME ); +if ( !(start_address &L3_MASK ) ) +if (need_pt_frame (start_address,L3_FRAME ) ) +new_pt_frame (&pt_pfn, pt_mfn, offset,L3_FRAME ); page = tab[offset]; - mfn =pte_to_mfn (page); - tab =to_virt (mfn_to_pfn (mfn) <<PAGE_SHIFT ); + pt_mfn =pte_to_mfn (page); + tab =to_virt (mfn_to_pfn (pt_mfn) <<PAGE_SHIFT );#endif offset =l3_table_offset (start_address);/* Need new L2 pt frame */ -if (!(start_address &L2_MASK )) -if (need_pt_frame (start_address,L2_FRAME )) -new_pt_frame (&pt_pfn, mfn, offset,L2_FRAME ); +if ( !(start_address &L2_MASK ) ) +if (need_pt_frame (start_address,L2_FRAME ) ) +new_pt_frame (&pt_pfn, pt_mfn, offset,L2_FRAME ); page = tab[offset]; - mfn =pte_to_mfn (page); - tab =to_virt (mfn_to_pfn (mfn) <<PAGE_SHIFT ); + pt_mfn =pte_to_mfn (page); + tab =to_virt (mfn_to_pfn (pt_mfn) <<PAGE_SHIFT ); offset =l2_table_offset (start_address);/* Need new L1 pt frame */ -if (!(start_address &L1_MASK )) -if (need_pt_frame (start_address,L1_FRAME )) -new_pt_frame (&pt_pfn, mfn, offset,L1_FRAME ); +if ( !(start_address &L1_MASK ) ) +if (need_pt_frame (start_address,L1_FRAME ) ) +new_pt_frame (&pt_pfn, pt_mfn, offset,L1_FRAME ); page = tab[offset]; - mfn =pte_to_mfn (page); + pt_mfn =pte_to_mfn (page); offset =l1_table_offset (start_address); - mmu_updates[count].ptr = ((pgentry_t )mfn <<PAGE_SHIFT ) +sizeof (pgentry_t) * offset; - mmu_updates[count].val = (pgentry_t )pfn_to_mfn (pfn_to_map++) <<PAGE_SHIFT |L1_PROT ; + mmu_updates[count].ptr = + ((pgentry_t )pt_mfn <<PAGE_SHIFT ) +sizeof (pgentry_t) * offset; + mmu_updates[count].val = + (pgentry_t )pfn_to_mfn (pfn_to_map++) <<PAGE_SHIFT |L1_PROT ; count++; -if (count ==L1_PAGETABLE_ENTRIES || pfn_to_map == *max_pfn) +if ( count ==L1_PAGETABLE_ENTRIES || pfn_to_map == *max_pfn ) { -if (HYPERVISOR_mmu_update (mmu_updates, count,NULL ,DOMID_SELF ) <0 ) + rc =HYPERVISOR_mmu_update (mmu_updates, count,NULL ,DOMID_SELF ); +if ( rc <0 ) { -printk ("PTE could not be updated\n" ); +printk ("ERROR: build_pagetable(): PTE could not be updated\n" ); +printk (" mmu_update failed with rc=%d\n" , rc);do_exit (); } count =0 ;
@@ -253,20 +278,26 @@ *start_pfn = pt_pfn; } +/* +* Mark portion of the address space read only. +*/ extern void shared_info;static void set_readonly (void *text,void *etext) { -unsigned long start_address = ((unsigned long ) text +PAGE_SIZE -1 ) &PAGE_MASK ; +unsigned long start_address = + ((unsigned long ) text +PAGE_SIZE -1 ) &PAGE_MASK ;unsigned long end_address = (unsigned long ) etext;static mmu_update_t mmu_updates[L1_PAGETABLE_ENTRIES +1 ];pgentry_t *tab = (pgentry_t *)start_info.pt_base, page;unsigned long mfn =pfn_to_mfn (virt_to_pfn (start_info.pt_base));unsigned long offset;int count =0 ; +int rc;printk ("setting %p-%p readonly\n" , text, etext); -while (start_address +PAGE_SIZE <= end_address) { +while ( start_address +PAGE_SIZE <= end_address ) + { tab = (pgentry_t *)start_info.pt_base; mfn =pfn_to_mfn (virt_to_pfn (start_info.pt_base));
@@ -287,20 +318,25 @@ offset =l1_table_offset (start_address); -if (start_address != (unsigned long )&shared_info) { - mmu_updates[count].ptr = ((pgentry_t )mfn <<PAGE_SHIFT ) +sizeof (pgentry_t) * offset; - mmu_updates[count].val = tab[offset] & ~_PAGE_RW; - count++; - }else -printk ("skipped %p\n" , start_address); +if ( start_address != (unsigned long )&shared_info ) + { + mmu_updates[count].ptr = + ((pgentry_t )mfn <<PAGE_SHIFT ) +sizeof (pgentry_t) * offset; + mmu_updates[count].val = tab[offset] & ~_PAGE_RW; + count++; + } +else +printk ("skipped %p\n" , start_address); start_address +=PAGE_SIZE ; -if (count ==L1_PAGETABLE_ENTRIES || start_address +PAGE_SIZE > end_address) +if ( count ==L1_PAGETABLE_ENTRIES || + start_address +PAGE_SIZE > end_address ) { -if (HYPERVISOR_mmu_update (mmu_updates, count,NULL ,DOMID_SELF ) <0 ) + rc =HYPERVISOR_mmu_update (mmu_updates, count,NULL ,DOMID_SELF ); +if ( rc <0 ) { -printk ("PTE could not be updated\n" ); +printk ("ERROR: set_readonly(): PTE could not be updated\n" );do_exit (); } count =0 ;
@@ -308,11 +344,11 @@ } { -mmuext_op_t op = { - .cmd =MMUEXT_TLB_FLUSH_ALL , - }; -int count; -HYPERVISOR_mmuext_op (&op,1 , &count,DOMID_SELF ); +mmuext_op_t op = { + .cmd =MMUEXT_TLB_FLUSH_ALL , + }; +int count; +HYPERVISOR_mmuext_op (&op,1 , &count,DOMID_SELF ); } }
@@ -371,7 +407,10 @@ } -static pgentry_t *get_pgt (unsigned long addr) +/* +* get the PTE for virtual address va if it exists. Otherwise NULL. +*/ +static pgentry_t *get_pgt (unsigned long va) {unsigned long mfn;pgentry_t *tab;
@@ -381,67 +420,78 @@ mfn =virt_to_mfn (start_info.pt_base);#if defined(__x86_64__) - offset =l4_table_offset (addr); -if (!(tab[offset] & _PAGE_PRESENT)) + offset =l4_table_offset (va); +if ( !(tab[offset] & _PAGE_PRESENT) )return NULL ; mfn =pte_to_mfn (tab[offset]); tab =mfn_to_virt (mfn);#endif - offset =l3_table_offset (addr); -if (!(tab[offset] & _PAGE_PRESENT)) + offset =l3_table_offset (va); +if ( !(tab[offset] & _PAGE_PRESENT) )return NULL ; mfn =pte_to_mfn (tab[offset]); tab =mfn_to_virt (mfn); - offset =l2_table_offset (addr); -if (!(tab[offset] & _PAGE_PRESENT)) + offset =l2_table_offset (va); +if ( !(tab[offset] & _PAGE_PRESENT) )return NULL ; mfn =pte_to_mfn (tab[offset]); tab =mfn_to_virt (mfn); - offset =l1_table_offset (addr); + offset =l1_table_offset (va);return &tab[offset]; } -pgentry_t *need_pgt (unsigned long addr) + +/* +* return a valid PTE for a given virtual address. If PTE does not exist, +* allocate page-table pages. +*/ +pgentry_t *need_pgt (unsigned long va) { -unsigned long mfn; +unsigned long pt_mfn;pgentry_t *tab;unsigned long pt_pfn;unsigned offset; tab = (pgentry_t *)start_info.pt_base; - mfn =virt_to_mfn (start_info.pt_base); + pt_mfn =virt_to_mfn (start_info.pt_base);#if defined(__x86_64__) - offset =l4_table_offset (addr); -if (!(tab[offset] & _PAGE_PRESENT)) { + offset =l4_table_offset (va); +if ( !(tab[offset] & _PAGE_PRESENT) ) + { pt_pfn =virt_to_pfn (alloc_page ()); -new_pt_frame (&pt_pfn, mfn, offset,L3_FRAME ); +new_pt_frame (&pt_pfn, pt_mfn, offset,L3_FRAME ); }ASSERT (tab[offset] & _PAGE_PRESENT); - mfn =pte_to_mfn (tab[offset]); - tab =mfn_to_virt (mfn); + pt_mfn =pte_to_mfn (tab[offset]); + tab =mfn_to_virt (pt_mfn);#endif - offset =l3_table_offset (addr); -if (!(tab[offset] & _PAGE_PRESENT)) { + offset =l3_table_offset (va); +if ( !(tab[offset] & _PAGE_PRESENT) ) + { pt_pfn =virt_to_pfn (alloc_page ()); -new_pt_frame (&pt_pfn, mfn, offset,L2_FRAME ); +new_pt_frame (&pt_pfn, pt_mfn, offset,L2_FRAME ); }ASSERT (tab[offset] & _PAGE_PRESENT); - mfn =pte_to_mfn (tab[offset]); - tab =mfn_to_virt (mfn); - offset =l2_table_offset (addr); -if (!(tab[offset] & _PAGE_PRESENT)) { + pt_mfn =pte_to_mfn (tab[offset]); + tab =mfn_to_virt (pt_mfn); + offset =l2_table_offset (va); +if ( !(tab[offset] & _PAGE_PRESENT) ) + { pt_pfn =virt_to_pfn (alloc_page ()); -new_pt_frame (&pt_pfn, mfn, offset,L1_FRAME ); +new_pt_frame (&pt_pfn, pt_mfn, offset,L1_FRAME ); }ASSERT (tab[offset] & _PAGE_PRESENT); - mfn =pte_to_mfn (tab[offset]); - tab =mfn_to_virt (mfn); + pt_mfn =pte_to_mfn (tab[offset]); + tab =mfn_to_virt (pt_mfn); - offset =l1_table_offset (addr); + offset =l1_table_offset (va);return &tab[offset]; } +/* +* Reserve an area of virtual address space for mappings and Heap +*/ static unsigned long demand_map_area_start;#ifdef __x86_64__#define DEMAND_MAP_PAGES ((128ULL << 30) / PAGE_SIZE)
@@ -466,7 +516,8 @@ demand_map_area_start = (unsigned long )pfn_to_virt (cur_pfn); cur_pfn +=DEMAND_MAP_PAGES ; -printk ("Demand map pfns at %lx-%lx.\n" , demand_map_area_start,pfn_to_virt (cur_pfn)); +printk ("Demand map pfns at %lx-%lx.\n" , + demand_map_area_start,pfn_to_virt (cur_pfn));#ifdef HAVE_LIBC cur_pfn++;
@@ -477,93 +528,123 @@#endif } +unsigned long allocate_ondemand (unsigned long n,unsigned long alignment) +{ +unsigned long x; +unsigned long y =0 ; + +/* Find a properly aligned run of n contiguous frames */ +for ( x =0 ; + x <=DEMAND_MAP_PAGES - n; + x = (x + y +1 + alignment -1 ) & ~(alignment -1 ) ) + { +unsigned long addr = demand_map_area_start + x *PAGE_SIZE ; +pgentry_t *pgt =get_pgt (addr); +for ( y =0 ; y < n; y++, addr +=PAGE_SIZE ) + { +if ( !(addr &L1_MASK ) ) + pgt =get_pgt (addr); +if ( pgt ) + { +if ( *pgt & _PAGE_PRESENT ) +break ; + pgt++; + } + } +if ( y == n ) +break ; + } +if ( y != n ) + { +printk ("Failed to find %ld frames!\n" , n); +return 0 ; + } +return demand_map_area_start + x *PAGE_SIZE ; +} + +/* +* Map an array of MFNs contiguously into virtual address space starting at +* va. map f[i*stride]+i*increment for i in 0..n-1. +*/ #define MAP_BATCH ((STACK_SIZE / 2) / sizeof(mmu_update_t)) -void do_map_frames (unsigned long addr, -unsigned long *f,unsigned long n,unsigned long stride, -unsigned long increment,domid_t id,int may_fail,unsigned long prot) +void do_map_frames (unsigned long va, +unsigned long *mfns,unsigned long n, +unsigned long stride,unsigned long incr, +domid_t id,int may_fail, +unsigned long prot) {pgentry_t *pgt =NULL ;unsigned long done =0 ;unsigned long i;int rc; -while (done < n) { -unsigned long todo; +if ( !mfns ) + { +printk ("do_map_frames: no mfns supplied\n" ); +return ; + } +DEBUG ("va=%p n=0x%lx, mfns[0]=0x%lx stride=0x%lx incr=0x%lx prot=0x%lx\n" , + va, n, mfns[0 ], stride, incr, prot); + +while ( done < n ) + { +unsigned long todo; -if (may_fail) - todo =1 ; -else - todo = n - done; +if ( may_fail ) + todo =1 ; +else + todo = n - done; -if (todo >MAP_BATCH ) - todo =MAP_BATCH ; +if ( todo >MAP_BATCH ) + todo =MAP_BATCH ; - { -mmu_update_t mmu_updates[todo]; + { +mmu_update_t mmu_updates[todo]; -for (i =0 ; i < todo; i++, addr +=PAGE_SIZE , pgt++) { -if (!pgt || !(addr &L1_MASK )) - pgt =need_pgt (addr); - mmu_updates[i].ptr =virt_to_mach (pgt); - mmu_updates[i].val = ((pgentry_t)(f[(done + i) * stride] + (done + i) * increment) <<PAGE_SHIFT ) | prot; - } +for ( i =0 ; i < todo; i++, va +=PAGE_SIZE , pgt++) + { +if ( !pgt || !(va &L1_MASK ) ) + pgt =need_pgt (va); + + mmu_updates[i].ptr =virt_to_mach (pgt) |MMU_NORMAL_PT_UPDATE ; + mmu_updates[i].val = ((pgentry_t)(mfns[(done + i) * stride] + + (done + i) * incr) + <<PAGE_SHIFT ) | prot; + } - rc =HYPERVISOR_mmu_update (mmu_updates, todo,NULL , id); -if (rc <0 ) { -if (may_fail) - f[done * stride] |=0xF0000000 ; -else { -printk ("Map %ld (%lx, ...) at %p failed: %d.\n" , todo, f[done * stride] + done * increment, addr, rc); + rc =HYPERVISOR_mmu_update (mmu_updates, todo,NULL , id); +if ( rc <0 ) + { +if (may_fail) + mfns[done * stride] |=0xF0000000 ; +else { +printk ("Map %ld (%lx, ...) at %p failed: %d.\n" , + todo, mfns[done * stride] + done * incr, va, rc);do_exit (); - } - } - } - - done += todo; + } + } + } + done += todo; } } -unsigned long allocate_ondemand (unsigned long n,unsigned long alignment) +/* +* Map an array of MFNs contiguous into virtual address space. Virtual +* addresses are allocated from the on demand area. +*/ +void *map_frames_ex (unsigned long *mfns,unsigned long n, +unsigned long stride,unsigned long incr, +unsigned long alignment, +domid_t id,int may_fail,unsigned long prot) { -unsigned long x; -unsigned long y =0 ; +unsigned long va =allocate_ondemand (n, alignment); -/* Find a properly aligned run of n contiguous frames */ -for (x =0 ; x <=DEMAND_MAP_PAGES - n; x = (x + y +1 + alignment -1 ) & ~(alignment -1 )) { -unsigned long addr = demand_map_area_start + x *PAGE_SIZE ; -pgentry_t *pgt =get_pgt (addr); -for (y =0 ; y < n; y++, addr +=PAGE_SIZE ) { -if (!(addr &L1_MASK )) - pgt =get_pgt (addr); -if (pgt) { -if (*pgt & _PAGE_PRESENT) -break ; - pgt++; - } - } -if (y == n) -break ; - } -if (y != n) { -printk ("Failed to find %ld frames!\n" , n); -return 0 ; - } -return demand_map_area_start + x *PAGE_SIZE ; -} - -void *map_frames_ex (unsigned long *f,unsigned long n,unsigned long stride, -unsigned long increment,unsigned long alignment,domid_t id, -int may_fail,unsigned long prot) -{ -unsigned long addr =allocate_ondemand (n, alignment); - -if (!addr) +if ( !va )return NULL ; -/* Found it at x. Map it in. */ -do_map_frames (addr, f, n, stride, increment, id, may_fail, prot); +do_map_frames (va, mfns, n, stride, incr, id, may_fail, prot); -return (void *)addr; +return (void *)va; }/*
@@ -632,15 +713,19 @@ } +/* +* Clear some of the bootstrap memory +*/ static void clear_bootstrap (void ) {pte_t nullpte = { }; +int rc;/* Use first page as the CoW zero page */ memset (&_text,0 ,PAGE_SIZE ); mfn_zero =virt_to_mfn ((unsigned long ) &_text); -if (HYPERVISOR_update_va_mapping (0 , nullpte,UVMF_INVLPG )) -printk ("Unable to unmap NULL page\n" ); +if ( (rc =HYPERVISOR_update_va_mapping (0 , nullpte,UVMF_INVLPG )) ) +printk ("Unable to unmap NULL page. rc=%d\n" , rc); }void arch_init_p2m (unsigned long max_pfn)
@@ -665,19 +750,19 @@unsigned long pfn; l3_list = (unsigned long *)alloc_page (); -for (pfn=0 ; pfn<max_pfn; pfn++) +for ( pfn=0 ; pfn<max_pfn; pfn++ ) { -if (!(pfn % (L1_P2M_ENTRIES *L2_P2M_ENTRIES ))) +if ( !(pfn % (L1_P2M_ENTRIES *L2_P2M_ENTRIES )) ) { l2_list = (unsigned long *)alloc_page (); -if ((pfn >>L3_P2M_SHIFT ) >0 ) +if ( (pfn >>L3_P2M_SHIFT ) >0 ) {printk ("Error: Too many pfns.\n" );do_exit (); } l3_list[(pfn >>L2_P2M_SHIFT )] =virt_to_mfn (l2_list); } -if (!(pfn % (L1_P2M_ENTRIES ))) +if ( !(pfn % (L1_P2M_ENTRIES )) ) { l1_list = (unsigned long *)alloc_page (); l2_list[(pfn >> L1_P2M_SHIFT) & L2_P2M_MASK] =
@@ -696,25 +781,25 @@unsigned long start_pfn, max_pfn, virt_pfns; -printk (" _text: %p\n" , &_text); -printk (" _etext: %p\n" , &_etext); -printk (" _erodata: %p\n" , &_erodata); -printk (" _edata: %p\n" , &_edata); -printk (" stack start: %p\n" , stack); -printk (" _end: %p\n" , &_end); +printk (" _text: %p(VA)\n" , &_text); +printk (" _etext: %p(VA)\n" , &_etext); +printk (" _erodata: %p(VA)\n" , &_erodata); +printk (" _edata: %p(VA)\n" , &_edata); +printk ("stack start: %p(VA)\n" , stack); +printk (" _end: %p(VA)\n" , &_end);/* First page follows page table pages and 3 more pages (store page etc) */ start_pfn =PFN_UP (to_phys (start_info.pt_base)) + - start_info.nr_pt_frames +3 ; + start_info.nr_pt_frames +3 ; max_pfn = start_info.nr_pages;/* We need room for demand mapping and heap, clip available memory */ virt_pfns =DEMAND_MAP_PAGES +HEAP_PAGES ; -if (max_pfn + virt_pfns +1 < max_pfn) +if ( max_pfn + virt_pfns +1 < max_pfn ) max_pfn = -(virt_pfns +1 ); -printk (" start_pfn: %lx\n" , start_pfn); -printk (" max_pfn: %lx\n" , max_pfn); +printk (" start_pfn: %lx\n" , start_pfn); +printk (" max_pfn: %lx\n" , max_pfn);build_pagetable (&start_pfn, &max_pfn);clear_bootstrap ();
extras/mini-os/arch/x86/setup.c +5 -3
@@ -63,10 +63,12 @@static shared_info_t *map_shared_info (unsigned long pa) { -if (HYPERVISOR_update_va_mapping ( - (unsigned long )shared_info,__pte (pa |7 ),UVMF_INVLPG ) ) +int rc; + +if ( (rc =HYPERVISOR_update_va_mapping ( + (unsigned long )shared_info,__pte (pa |7 ),UVMF_INVLPG )) ) { -printk ("Failed to map shared_info!!\n" ); +printk ("Failed to map shared_info!! rc=%d\n" , rc);do_exit (); }return (shared_info_t *)shared_info;
extras/mini-os/kernel.c +10 -8
@@ -490,14 +490,16 @@/* print out some useful information */ printk ("Xen Minimal OS!\n" ); -printk ("start_info: %p\n" , si); -printk (" nr_pages: %lu" , si->nr_pages); -printk (" shared_inf: %08lx\n" , si->shared_info); -printk (" pt_base: %p" , (void *)si->pt_base); -printk (" mod_start: 0x%lx\n" , si->mod_start); -printk (" mod_len: %lu\n" , si->mod_len); -printk (" flags: 0x%x\n" , (unsigned int )si->flags); -printk (" cmd_line: %s\n" , +printk (" start_info: %p(VA)\n" , si); +printk (" nr_pages: 0x%lx\n" , si->nr_pages); +printk (" shared_inf: 0x%08lx(MA)\n" , si->shared_info); +printk (" pt_base: %p(VA)\n" , (void *)si->pt_base); +printk ("nr_pt_frames: 0x%lx\n" , si->nr_pt_frames); +printk (" mfn_list: %p(VA)\n" , (void *)si->mfn_list); +printk (" mod_start: 0x%lx(VA)\n" , si->mod_start); +printk (" mod_len: %lu\n" , si->mod_len); +printk (" flags: 0x%x\n" , (unsigned int )si->flags); +printk (" cmd_line: %s\n" , si->cmd_line ? (const char *)si->cmd_line :"NULL" );/* Set up events. */
_______________________________________________
Xen-devel mailing list
[email protected]
http://lists.xensource.com/xen-devel