diff options
Diffstat (limited to 'tools')
| -rw-r--r-- | tools/Makefile | 4 | ||||
| -rwxr-xr-x | tools/examples/xc_dom_control.py | 4 | ||||
| -rw-r--r-- | tools/xc/lib/Makefile | 2 | ||||
| -rw-r--r-- | tools/xc/lib/xc_domain.c | 1 | ||||
| -rw-r--r-- | tools/xc/lib/xc_linux_build.c | 25 | ||||
| -rw-r--r-- | tools/xc/lib/xc_linux_restore.c | 242 | ||||
| -rw-r--r-- | tools/xc/lib/xc_linux_save.c | 392 | ||||
| -rw-r--r-- | tools/xc/lib/xc_netbsd_build.c | 26 | ||||
| -rw-r--r-- | tools/xc/lib/xc_private.c | 228 | ||||
| -rw-r--r-- | tools/xc/lib/xc_private.h | 100 |
10 files changed, 787 insertions, 237 deletions
diff --git a/tools/Makefile b/tools/Makefile index 0d4c43fb24..9ddf5f25a2 100644 --- a/tools/Makefile +++ b/tools/Makefile @@ -17,6 +17,10 @@ install: all $(MAKE) -C xenctl install $(MAKE) -C xend install +dist: $(TARGET) + $(MAKE) prefix=`pwd`/../../install dist=yes install + + clean: $(MAKE) -C balloon clean $(MAKE) -C xc clean diff --git a/tools/examples/xc_dom_control.py b/tools/examples/xc_dom_control.py index ea97a45f0e..4f0bd5de52 100755 --- a/tools/examples/xc_dom_control.py +++ b/tools/examples/xc_dom_control.py @@ -137,6 +137,10 @@ elif cmd == 'suspend': os.kill(pid, signal.SIGTERM) xc.domain_stop( dom=dom ) + + while not xc.domain_getinfo( first_dom=dom, max_doms=1 )[0]['stopped']: + time.sleep(0.1); + rc = xc.linux_save( dom=dom, state_file=file, progress=1) if rc == 0 : xc.domain_destroy( dom=dom, force=1 ) diff --git a/tools/xc/lib/Makefile b/tools/xc/lib/Makefile index 79dce046df..f542935167 100644 --- a/tools/xc/lib/Makefile +++ b/tools/xc/lib/Makefile @@ -4,7 +4,7 @@ MINOR = 0 SONAME = libxc.so.$(MAJOR) CC = gcc -CFLAGS = -c -Wall -O3 -fno-strict-aliasing +CFLAGS = -c -Werror -O3 -fno-strict-aliasing CFLAGS += -I../../../xen/include/hypervisor-ifs CFLAGS += -I../../xend/lib CFLAGS += -I../../../xenolinux-sparse/include diff --git a/tools/xc/lib/xc_domain.c b/tools/xc/lib/xc_domain.c index ec28f2686b..1d77bfc016 100644 --- a/tools/xc/lib/xc_domain.c +++ b/tools/xc/lib/xc_domain.c @@ -84,6 +84,7 @@ int xc_domain_getinfo(int xc_handle, { op.cmd = DOM0_GETDOMAININFO; op.u.getdomaininfo.domain = (domid_t)next_domid; + op.u.getdomaininfo.ctxt = NULL; // no exec context info, thanks. if ( do_dom0_op(xc_handle, &op) < 0 ) break; info->domid = (u64)op.u.getdomaininfo.domain; diff --git a/tools/xc/lib/xc_linux_build.c b/tools/xc/lib/xc_linux_build.c index 27bc6c6668..67351210fc 100644 --- a/tools/xc/lib/xc_linux_build.c +++ b/tools/xc/lib/xc_linux_build.c @@ -26,6 +26,7 @@ static long get_tot_pages(int xc_handle, u64 domid) dom0_op_t op; op.cmd = DOM0_GETDOMAININFO; op.u.getdomaininfo.domain = (domid_t)domid; + op.u.getdomaininfo.ctxt = NULL; return (do_dom0_op(xc_handle, &op) < 0) ? -1 : op.u.getdomaininfo.tot_pages; } @@ -70,7 +71,7 @@ static int setup_guestos(int xc_handle, gzFile initrd_gfd, unsigned long initrd_len, unsigned long nr_pages, unsigned long *pvsi, unsigned long *pvke, - dom0_builddomain_t *builddomain, + full_execution_context_t *ctxt, const char *cmdline, unsigned long shared_info_frame, unsigned int control_evtchn, @@ -164,8 +165,6 @@ static int setup_guestos(int xc_handle, v_start, v_end); printf(" ENTRY ADDRESS: %08lx\n", vkern_entry); - memset(builddomain, 0, sizeof(*builddomain)); - if ( (pm_handle = init_pfn_mapper((domid_t)dom)) < 0 ) goto error_out; @@ -206,7 +205,7 @@ static int setup_guestos(int xc_handle, /* First allocate page for page dir. */ ppt_alloc = (vpt_start - v_start) >> PAGE_SHIFT; l2tab = page_array[ppt_alloc++] << PAGE_SHIFT; - builddomain->ctxt.pt_base = l2tab; + ctxt->pt_base = l2tab; /* Initialise the page tables. */ if ( (vl2tab = map_pfn_writeable(pm_handle, l2tab >> PAGE_SHIFT)) == NULL ) @@ -390,7 +389,7 @@ int xc_linux_build(int xc_handle, int initrd_fd = -1; gzFile initrd_gfd = NULL; int rc, i; - full_execution_context_t *ctxt; + full_execution_context_t st_ctxt, *ctxt = &st_ctxt; unsigned long nr_pages; char *image = NULL; unsigned long image_size, initrd_size=0; @@ -422,8 +421,15 @@ int xc_linux_build(int xc_handle, } } + if ( mlock(&st_ctxt, sizeof(st_ctxt) ) ) + { + PERROR("Unable to mlock ctxt"); + return 1; + } + op.cmd = DOM0_GETDOMAININFO; op.u.getdomaininfo.domain = (domid_t)domid; + op.u.getdomaininfo.ctxt = ctxt; if ( (do_dom0_op(xc_handle, &op) < 0) || ((u64)op.u.getdomaininfo.domain != domid) ) { @@ -431,7 +437,7 @@ int xc_linux_build(int xc_handle, goto error_out; } if ( (op.u.getdomaininfo.state != DOMSTATE_STOPPED) || - (op.u.getdomaininfo.ctxt.pt_base != 0) ) + (ctxt->pt_base != 0) ) { ERROR("Domain is already constructed"); goto error_out; @@ -440,7 +446,7 @@ int xc_linux_build(int xc_handle, if ( setup_guestos(xc_handle, domid, image, image_size, initrd_gfd, initrd_size, nr_pages, &vstartinfo_start, &vkern_entry, - &launch_op.u.builddomain, cmdline, + ctxt, cmdline, op.u.getdomaininfo.shared_info_frame, control_evtchn, io_priv) < 0 ) { @@ -455,8 +461,6 @@ int xc_linux_build(int xc_handle, if ( image != NULL ) free(image); - ctxt = &launch_op.u.builddomain.ctxt; - ctxt->flags = 0; /* @@ -509,8 +513,11 @@ int xc_linux_build(int xc_handle, ctxt->failsafe_callback_cs = FLAT_GUESTOS_CS; ctxt->failsafe_callback_eip = 0; + memset( &launch_op, 0, sizeof(launch_op) ); + launch_op.u.builddomain.domain = (domid_t)domid; launch_op.u.builddomain.num_vifs = 1; + launch_op.u.builddomain.ctxt = ctxt; launch_op.cmd = DOM0_BUILDDOMAIN; rc = do_dom0_op(xc_handle, &launch_op); diff --git a/tools/xc/lib/xc_linux_restore.c b/tools/xc/lib/xc_linux_restore.c index 239df65984..e27221281a 100644 --- a/tools/xc/lib/xc_linux_restore.c +++ b/tools/xc/lib/xc_linux_restore.c @@ -10,6 +10,8 @@ #include <asm-xen/suspend.h> #include <zlib.h> +#define MAX_BATCH_SIZE 1024 + /* This may allow us to create a 'quiet' command-line option, if necessary. */ #define verbose_printf(_f, _a...) \ do { \ @@ -58,8 +60,8 @@ int xc_linux_restore(int xc_handle, u64 *pdomid) { dom0_op_t op; - int rc = 1, i, j; - unsigned long mfn, pfn; + int rc = 1, i, j, n, k; + unsigned long mfn, pfn, xpfn; unsigned int prev_pc, this_pc; /* Number of page frames in use by this Linux session. */ @@ -93,6 +95,9 @@ int xc_linux_restore(int xc_handle, /* A temporary mapping of the guest's suspend record. */ suspend_record_t *p_srec; + mfn_mapper_t *region_mapper, *mapper_handle1; + char *region_base; + /* The name and descriptor of the file that we are reading from. */ int fd; gzFile gfd; @@ -114,6 +119,14 @@ int xc_linux_restore(int xc_handle, return 1; } + if ( mlock(&ctxt, sizeof(ctxt) ) ) + { + /* needed for when we do the build dom0 op, + but might as well do early */ + PERROR("Unable to mlock ctxt"); + return 1; + } + /* Start writing out the saved-domain record. */ if ( !checked_read(gfd, signature, 16) || (memcmp(signature, "LinuxGuestRecord", 16) != 0) ) @@ -159,12 +172,6 @@ int xc_linux_restore(int xc_handle, goto out; } - if ( !checked_read(gfd, pfn_type, 4 * nr_pfns) ) - { - ERROR("Error when reading from state file"); - goto out; - } - /* Set the domain's name to that from the restore file */ if ( xc_domain_setname( xc_handle, dom, name ) ) { @@ -184,6 +191,7 @@ int xc_linux_restore(int xc_handle, /* Get the domain's shared-info frame. */ op.cmd = DOM0_GETDOMAININFO; op.u.getdomaininfo.domain = (domid_t)dom; + op.u.getdomaininfo.ctxt = NULL; if ( do_dom0_op(xc_handle, &op) < 0 ) { ERROR("Could not get information on new domain"); @@ -212,6 +220,15 @@ int xc_linux_restore(int xc_handle, goto out; } + + if ( (region_mapper = mfn_mapper_init(xc_handle, dom, + MAX_BATCH_SIZE*PAGE_SIZE, + PROT_WRITE )) + == NULL ) + goto out; + + region_base = mfn_mapper_base( region_mapper ); + verbose_printf("Reloading memory pages: 0%%"); /* @@ -219,75 +236,141 @@ int xc_linux_restore(int xc_handle, * We uncanonicalise page tables as we go. */ prev_pc = 0; - for ( i = 0; i < nr_pfns; i++ ) + + n=0; + while(1) { - this_pc = (i * 100) / nr_pfns; + int j; + unsigned long region_pfn_type[1024]; + + this_pc = (n * 100) / nr_pfns; if ( (this_pc - prev_pc) >= 5 ) { verbose_printf("\b\b\b\b%3d%%", this_pc); prev_pc = this_pc; } - mfn = pfn_to_mfn_table[i]; - - ppage = map_pfn_writeable(pm_handle, mfn); - - if ( !checked_read(gfd, ppage, PAGE_SIZE) ) + if ( !checked_read(gfd, &j, sizeof(int)) ) { ERROR("Error when reading from state file"); goto out; } - if ( pfn_type[i] == L1TAB ) + //printf("batch=%d\n",j); + + if(j==0) break; // our work here is done + + if ( !checked_read(gfd, region_pfn_type, j*sizeof(unsigned long)) ) { - for ( j = 0; j < 1024; j++ ) - { - if ( ppage[j] & _PAGE_PRESENT ) - { - if ( (pfn = ppage[j] >> PAGE_SHIFT) >= nr_pfns ) - { - ERROR("Frame number in page table is out of range"); - goto out; - } - if ( (pfn_type[pfn] != NONE) && (ppage[j] & _PAGE_RW) ) - { - ERROR("Write access requested for a restricted frame"); - goto out; - } - ppage[j] &= (PAGE_SIZE - 1) & ~(_PAGE_GLOBAL | _PAGE_PAT); - ppage[j] |= pfn_to_mfn_table[pfn] << PAGE_SHIFT; - } - } - } - else if ( pfn_type[i] == L2TAB ) - { - for ( j = 0; j < (HYPERVISOR_VIRT_START>>L2_PAGETABLE_SHIFT); j++ ) - { - if ( ppage[j] & _PAGE_PRESENT ) - { - if ( (pfn = ppage[j] >> PAGE_SHIFT) >= nr_pfns ) - { - ERROR("Frame number in page table is out of range"); - goto out; - } - if ( pfn_type[pfn] != L1TAB ) - { - ERROR("Page table mistyping"); - goto out; - } - ppage[j] &= (PAGE_SIZE - 1) & ~(_PAGE_GLOBAL | _PAGE_PSE); - ppage[j] |= pfn_to_mfn_table[pfn] << PAGE_SHIFT; - } - } + ERROR("Error when reading from state file"); + goto out; } - unmap_pfn(pm_handle, ppage); + for(i=0;i<j;i++) + { + pfn = region_pfn_type[i] & ~PGT_type_mask; + mfn = pfn_to_mfn_table[pfn]; + + mfn_mapper_queue_entry( region_mapper, i<<PAGE_SHIFT, + mfn, PAGE_SIZE ); + } + + if( mfn_mapper_flush_queue(region_mapper) ) + { + ERROR("Couldn't map page region"); + goto out; + } + + + for(i=0;i<j;i++) + { + unsigned long *ppage; + + pfn = region_pfn_type[i] & ~PGT_type_mask; + +//if(pfn_type[i])printf("^pfn=%d %08lx\n",pfn,pfn_type[i]); + + if (pfn>nr_pfns) + { + ERROR("pfn out of range"); + goto out; + } + + region_pfn_type[i] &= PGT_type_mask; + + pfn_type[pfn] = region_pfn_type[i]; + + mfn = pfn_to_mfn_table[pfn]; + +//if(region_pfn_type[i])printf("i=%d pfn=%d mfn=%d type=%lx\n",i,pfn,mfn,region_pfn_type[i]); + + ppage = (unsigned long*) (region_base + i*PAGE_SIZE); + + if ( !checked_read(gfd, ppage, PAGE_SIZE) ) + { + ERROR("Error when reading from state file"); + goto out; + } + + if ( region_pfn_type[i] == L1TAB ) + { + for ( k = 0; k < 1024; k++ ) + { + if ( ppage[k] & _PAGE_PRESENT ) + { + if ( (xpfn = ppage[k] >> PAGE_SHIFT) >= nr_pfns ) + { + ERROR("Frame number in type %d page table is out of range. i=%d k=%d pfn=%d nr_pfns=%d",region_pfn_type[i],i,k,xpfn,nr_pfns); + goto out; + } +#if 0 + if ( (region_pfn_type[xpfn] != NONE) && (ppage[k] & _PAGE_RW) ) + { + ERROR("Write access requested for a restricted frame"); + goto out; + } +#endif + ppage[k] &= (PAGE_SIZE - 1) & ~(_PAGE_GLOBAL | _PAGE_PAT); + ppage[k] |= pfn_to_mfn_table[xpfn] << PAGE_SHIFT; + } + } + } + else if ( region_pfn_type[i] == L2TAB ) + { + for ( k = 0; k < (HYPERVISOR_VIRT_START>>L2_PAGETABLE_SHIFT); k++ ) + { + if ( ppage[k] & _PAGE_PRESENT ) + { + if ( (xpfn = ppage[k] >> PAGE_SHIFT) >= nr_pfns ) + { + ERROR("Frame number in page table is out of range"); + goto out; + } +#if 0 + if ( region_pfn_type[pfn] != L1TAB ) + { + ERROR("Page table mistyping"); + goto out; + } +#endif + ppage[k] &= (PAGE_SIZE - 1) & ~(_PAGE_GLOBAL | _PAGE_PSE); + ppage[k] |= pfn_to_mfn_table[xpfn] << PAGE_SHIFT; + } + } + } + + if ( add_mmu_update(xc_handle, mmu, + (mfn<<PAGE_SHIFT) | MMU_MACHPHYS_UPDATE, pfn) ) + goto out; + + } + + n+=j; // crude stats - if ( add_mmu_update(xc_handle, mmu, - (mfn<<PAGE_SHIFT) | MMU_MACHPHYS_UPDATE, i) ) - goto out; } + mfn_mapper_close( region_mapper ); + /* * Pin page tables. Do this after writing to them as otherwise Xen * will barf when doing the type-checking. @@ -352,26 +435,47 @@ int xc_linux_restore(int xc_handle, pfn = ctxt.pt_base >> PAGE_SHIFT; if ( (pfn >= nr_pfns) || (pfn_type[pfn] != L2TAB) ) { - ERROR("PT base is bad"); + printf("PT base is bad. pfn=%d nr=%d type=%08lx %08lx\n", + pfn, nr_pfns, pfn_type[pfn], L2TAB); + ERROR("PT base is bad."); goto out; } ctxt.pt_base = pfn_to_mfn_table[pfn] << PAGE_SHIFT; /* Uncanonicalise the pfn-to-mfn table frame-number list. */ - for ( i = 0; i < nr_pfns; i += 1024 ) + + + if ( (mapper_handle1 = mfn_mapper_init(xc_handle, dom, + 1024*1024, PROT_WRITE )) + == NULL ) + goto out; + + for ( i = 0; i < (nr_pfns+1023)/1024; i++ ) { - unsigned long copy_size = (nr_pfns - i) * sizeof(unsigned long); - if ( copy_size > PAGE_SIZE ) copy_size = PAGE_SIZE; - pfn = pfn_to_mfn_frame_list[i/1024]; + unsigned long pfn, mfn; + + pfn = pfn_to_mfn_frame_list[i]; if ( (pfn >= nr_pfns) || (pfn_type[pfn] != NONE) ) { ERROR("PFN-to-MFN frame number is bad"); goto out; } - ppage = map_pfn_writeable(pm_handle, pfn_to_mfn_table[pfn]); - memcpy(ppage, &pfn_to_mfn_table[i], copy_size); - unmap_pfn(pm_handle, ppage); + mfn = pfn_to_mfn_table[pfn]; + + mfn_mapper_queue_entry( mapper_handle1, i<<PAGE_SHIFT, + mfn, PAGE_SIZE ); } + + if ( mfn_mapper_flush_queue(mapper_handle1) ) + { + ERROR("Couldn't map pfn_to_mfn table"); + goto out; + } + + memcpy( mfn_mapper_base( mapper_handle1 ), pfn_to_mfn_table, + nr_pfns*sizeof(unsigned long) ); + + mfn_mapper_close( mapper_handle1 ); /* * Safety checking of saved context: @@ -406,11 +510,11 @@ int xc_linux_restore(int xc_handle, ERROR("Bad LDT base or size"); goto out; } - + op.cmd = DOM0_BUILDDOMAIN; op.u.builddomain.domain = (domid_t)dom; op.u.builddomain.num_vifs = 1; - memcpy(&op.u.builddomain.ctxt, &ctxt, sizeof(ctxt)); + op.u.builddomain.ctxt = &ctxt; rc = do_dom0_op(xc_handle, &op); out: diff --git a/tools/xc/lib/xc_linux_save.c b/tools/xc/lib/xc_linux_save.c index dc759f546c..88ed9e15d7 100644 --- a/tools/xc/lib/xc_linux_save.c +++ b/tools/xc/lib/xc_linux_save.c @@ -10,6 +10,8 @@ #include <asm-xen/suspend.h> #include <zlib.h> +#define BATCH_SIZE 1024 /* 1024 pages (4MB) at a time */ + /* This may allow us to create a 'quiet' command-line option, if necessary. */ #define verbose_printf(_f, _a...) \ do { \ @@ -24,7 +26,7 @@ */ #define MFN_IS_IN_PSEUDOPHYS_MAP(_mfn) \ (((_mfn) < (1024*1024)) && \ - (pfn_to_mfn_table[mfn_to_pfn_table[_mfn]] == (_mfn))) + (live_pfn_to_mfn_table[live_mfn_to_pfn_table[_mfn]] == (_mfn))) /* Returns TRUE if MFN is successfully converted to a PFN. */ #define translate_mfn_to_pfn(_pmfn) \ @@ -34,37 +36,11 @@ if ( !MFN_IS_IN_PSEUDOPHYS_MAP(mfn) ) \ _res = 0; \ else \ - *(_pmfn) = mfn_to_pfn_table[mfn]; \ + *(_pmfn) = live_mfn_to_pfn_table[mfn]; \ _res; \ }) -static int check_pfn_ownership(int xc_handle, - unsigned long mfn, - u64 dom) -{ - dom0_op_t op; - op.cmd = DOM0_GETPAGEFRAMEINFO; - op.u.getpageframeinfo.pfn = mfn; - op.u.getpageframeinfo.domain = (domid_t)dom; - return (do_dom0_op(xc_handle, &op) >= 0); -} -#define GETPFN_ERR (~0U) -static unsigned int get_pfn_type(int xc_handle, - unsigned long mfn, - u64 dom) -{ - dom0_op_t op; - op.cmd = DOM0_GETPAGEFRAMEINFO; - op.u.getpageframeinfo.pfn = mfn; - op.u.getpageframeinfo.domain = (domid_t)dom; - if ( do_dom0_op(xc_handle, &op) < 0 ) - { - PERROR("Unexpected failure when getting page frame info!"); - return GETPFN_ERR; - } - return op.u.getpageframeinfo.type; -} static int checked_write(gzFile fd, void *buf, size_t count) { @@ -80,10 +56,13 @@ int xc_linux_save(int xc_handle, int verbose) { dom0_op_t op; - int rc = 1, i, j; + int rc = 1, i, j, k, n; unsigned long mfn; unsigned int prev_pc, this_pc; + /* state of the new MFN mapper */ + mfn_mapper_t *mapper_handle1, *mapper_handle2; + /* Remember if we stopped the guest, so we can restart it on exit. */ int we_stopped_it = 0; @@ -100,18 +79,23 @@ int xc_linux_save(int xc_handle, unsigned long *pfn_type = NULL; /* A temporary mapping, and a copy, of one frame of guest memory. */ - unsigned long *ppage, page[1024]; + unsigned long page[1024]; - /* A temporary mapping, and a copy, of the pfn-to-mfn table frame list. */ - unsigned long *p_pfn_to_mfn_frame_list, pfn_to_mfn_frame_list[1024]; - /* A temporary mapping of one frame in the above list. */ - unsigned long *pfn_to_mfn_frame; + /* A copy of the pfn-to-mfn table frame list. */ + unsigned long *live_pfn_to_mfn_frame_list; + unsigned long pfn_to_mfn_frame_list[1024]; - /* A table mapping each PFN to its current MFN. */ - unsigned long *pfn_to_mfn_table = NULL; - /* A table mapping each current MFN to its canonical PFN. */ - unsigned long *mfn_to_pfn_table = NULL; + /* Live mapping of the table mapping each PFN to its current MFN. */ + unsigned long *live_pfn_to_mfn_table = NULL; + /* Live mapping of system MFN to PFN table. */ + unsigned long *live_mfn_to_pfn_table = NULL; + /* Live mapping of shared info structure */ + unsigned long *live_shinfo; + + /* base of the region in which domain memory is mapped */ + unsigned char *region_base; + /* A temporary mapping, and a copy, of the guest's suspend record. */ suspend_record_t *p_srec, srec; @@ -138,11 +122,18 @@ int xc_linux_save(int xc_handle, return 1; } + if ( mlock(&ctxt, sizeof(ctxt) ) ) + { + PERROR("Unable to mlock ctxt"); + return 1; + } + /* Ensure that the domain exists, and that it is stopped. */ for ( ; ; ) { op.cmd = DOM0_GETDOMAININFO; op.u.getdomaininfo.domain = (domid_t)domid; + op.u.getdomaininfo.ctxt = &ctxt; if ( (do_dom0_op(xc_handle, &op) < 0) || ((u64)op.u.getdomaininfo.domain != domid) ) { @@ -150,7 +141,6 @@ int xc_linux_save(int xc_handle, goto out; } - memcpy(&ctxt, &op.u.getdomaininfo.ctxt, sizeof(ctxt)); memcpy(name, op.u.getdomaininfo.name, sizeof(name)); shared_info_frame = op.u.getdomaininfo.shared_info_frame; @@ -178,99 +168,115 @@ int xc_linux_save(int xc_handle, goto out; } - if ( (pm_handle = init_pfn_mapper((domid_t)domid)) < 0 ) - goto out; - /* Is the suspend-record MFN actually valid for this domain? */ - if ( !check_pfn_ownership(xc_handle, ctxt.cpu_ctxt.esi, domid) ) + /* Map the suspend-record MFN to pin it. The page must be owned by + domid for this to succeed. */ + p_srec = mfn_mapper_map_single(xc_handle, domid, + sizeof(srec), PROT_READ, + ctxt.cpu_ctxt.esi ); + + if (!p_srec) { - ERROR("Invalid state record pointer"); + ERROR("Couldn't map state record"); goto out; } - /* If the suspend-record MFN is okay then grab a copy of it to @srec. */ - p_srec = map_pfn_readonly(pm_handle, ctxt.cpu_ctxt.esi); - memcpy(&srec, p_srec, sizeof(srec)); - unmap_pfn(pm_handle, p_srec); + memcpy( &srec, p_srec, sizeof(srec) ); + /* cheesy sanity check */ if ( srec.nr_pfns > 1024*1024 ) { ERROR("Invalid state record -- pfn count out of range"); goto out; } - if ( !check_pfn_ownership(xc_handle, srec.pfn_to_mfn_frame_list, domid) ) + /* the pfn_to_mfn_frame_list fits in a single page */ + live_pfn_to_mfn_frame_list = + mfn_mapper_map_single(xc_handle, domid, + PAGE_SIZE, PROT_READ, + srec.pfn_to_mfn_frame_list ); + + if (!live_pfn_to_mfn_frame_list) { - ERROR("Invalid pfn-to-mfn frame list pointer"); + ERROR("Couldn't map pfn_to_mfn_frame_list"); goto out; } + + + if ( (mapper_handle1 = mfn_mapper_init(xc_handle, domid, + 1024*1024, PROT_READ )) + == NULL ) + goto out; + + for ( i = 0; i < (srec.nr_pfns+1023)/1024; i++ ) + { + /* Grab a copy of the pfn-to-mfn table frame list. + This has the effect of preventing the page from being freed and + given to another domain. (though the domain is stopped anyway...) */ + mfn_mapper_queue_entry( mapper_handle1, i<<PAGE_SHIFT, + live_pfn_to_mfn_frame_list[i], + PAGE_SIZE ); + } + + if ( mfn_mapper_flush_queue(mapper_handle1) ) + { + ERROR("Couldn't map pfn_to_mfn table"); + goto out; + } + + live_pfn_to_mfn_table = mfn_mapper_base( mapper_handle1 ); + - /* Grab a copy of the pfn-to-mfn table frame list. */ - p_pfn_to_mfn_frame_list = map_pfn_readonly( - pm_handle, srec.pfn_to_mfn_frame_list); - memcpy(pfn_to_mfn_frame_list, p_pfn_to_mfn_frame_list, PAGE_SIZE); - unmap_pfn(pm_handle, p_pfn_to_mfn_frame_list); /* We want zeroed memory so use calloc rather than malloc. */ - mfn_to_pfn_table = calloc(1, 4 * 1024 * 1024); - pfn_to_mfn_table = calloc(1, 4 * srec.nr_pfns); - pfn_type = calloc(1, 4 * srec.nr_pfns); + pfn_type = calloc(BATCH_SIZE, sizeof(unsigned long)); - if ( (mfn_to_pfn_table == NULL) || - (pfn_to_mfn_table == NULL) || - (pfn_type == NULL) ) + if ( (pfn_type == NULL) ) { errno = ENOMEM; goto out; } + if ( mlock( pfn_type, BATCH_SIZE * sizeof(unsigned long) ) ) + { + ERROR("Unable to mlock"); + goto out; + } - /* - * Construct the local pfn-to-mfn and mfn-to-pfn tables. On exit from this - * loop we have each MFN mapped at most once. Note that there may be MFNs - * that aren't mapped at all: we detect these by MFN_IS_IN_PSEUDOPHYS_MAP. - */ - pfn_to_mfn_frame = NULL; - for ( i = 0; i < srec.nr_pfns; i++ ) + + /* Track the mfn_to_pfn table down from the domains PT */ { - /* Each frameful of table frames must be checked & mapped on demand. */ - if ( (i & 1023) == 0 ) - { - mfn = pfn_to_mfn_frame_list[i/1024]; - if ( !check_pfn_ownership(xc_handle, mfn, domid) ) - { - ERROR("Invalid frame number if pfn-to-mfn frame list"); - goto out; - } - if ( pfn_to_mfn_frame != NULL ) - unmap_pfn(pm_handle, pfn_to_mfn_frame); - pfn_to_mfn_frame = map_pfn_readonly(pm_handle, mfn); - } - - mfn = pfn_to_mfn_frame[i & 1023]; + unsigned long *pgd; + unsigned long mfn_to_pfn_table_start_mfn; - if ( !check_pfn_ownership(xc_handle, mfn, domid) ) - { - ERROR("Invalid frame specified with pfn-to-mfn table"); - goto out; - } + pgd = mfn_mapper_map_single(xc_handle, domid, + PAGE_SIZE, PROT_READ, + ctxt.pt_base>>PAGE_SHIFT); - /* Did we map this MFN already? That would be invalid! */ - if ( MFN_IS_IN_PSEUDOPHYS_MAP(mfn) ) - { - ERROR("A machine frame appears twice in pseudophys space"); - goto out; - } + mfn_to_pfn_table_start_mfn = + pgd[HYPERVISOR_VIRT_START>>L2_PAGETABLE_SHIFT]>>PAGE_SHIFT; + + live_mfn_to_pfn_table = + mfn_mapper_map_single(xc_handle, ~0ULL, + PAGE_SIZE*1024, PROT_READ, + mfn_to_pfn_table_start_mfn ); + } - pfn_to_mfn_table[i] = mfn; - mfn_to_pfn_table[mfn] = i; - /* Query page type by MFN, but store it by PFN. */ - if ( (pfn_type[i] = get_pfn_type(xc_handle, mfn, domid)) == - GETPFN_ERR ) - goto out; + /* + * Quick belt and braces sanity check. + */ + + for ( i = 0; i < srec.nr_pfns; i++ ) + { + mfn = live_pfn_to_mfn_table[i]; + + if( live_mfn_to_pfn_table[mfn] != i ) + printf("i=%d mfn=%d live_mfn_to_pfn_table=%d\n", + i,mfn,live_mfn_to_pfn_table[mfn]); } + /* Canonicalise the suspend-record frame number. */ if ( !translate_mfn_to_pfn(&ctxt.cpu_ctxt.esi) ) { @@ -294,9 +300,10 @@ int xc_linux_save(int xc_handle, ERROR("PT base is not in range of pseudophys map"); goto out; } - ctxt.pt_base = mfn_to_pfn_table[ctxt.pt_base >> PAGE_SHIFT] << PAGE_SHIFT; + ctxt.pt_base = live_mfn_to_pfn_table[ctxt.pt_base >> PAGE_SHIFT] << PAGE_SHIFT; /* Canonicalise the pfn-to-mfn table frame-number list. */ + memcpy( pfn_to_mfn_frame_list, live_pfn_to_mfn_frame_list, PAGE_SIZE ); for ( i = 0; i < srec.nr_pfns; i += 1024 ) { if ( !translate_mfn_to_pfn(&pfn_to_mfn_frame_list[i/1024]) ) @@ -307,63 +314,152 @@ int xc_linux_save(int xc_handle, } /* Start writing out the saved-domain record. */ - ppage = map_pfn_readonly(pm_handle, shared_info_frame); + live_shinfo = mfn_mapper_map_single(xc_handle, domid, + PAGE_SIZE, PROT_READ, + shared_info_frame); + + if (!live_shinfo) + { + ERROR("Couldn't map live_shinfo"); + goto out; + } + if ( !checked_write(gfd, "LinuxGuestRecord", 16) || !checked_write(gfd, name, sizeof(name)) || !checked_write(gfd, &srec.nr_pfns, sizeof(unsigned long)) || !checked_write(gfd, &ctxt, sizeof(ctxt)) || - !checked_write(gfd, ppage, PAGE_SIZE) || - !checked_write(gfd, pfn_to_mfn_frame_list, PAGE_SIZE) || - !checked_write(gfd, pfn_type, 4 * srec.nr_pfns) ) + !checked_write(gfd, live_shinfo, PAGE_SIZE) || + !checked_write(gfd, pfn_to_mfn_frame_list, PAGE_SIZE) ) { ERROR("Error when writing to state file"); goto out; } - unmap_pfn(pm_handle, ppage); + munmap(live_shinfo, PAGE_SIZE); verbose_printf("Saving memory pages: 0%%"); + if ( (mapper_handle2 = mfn_mapper_init(xc_handle, domid, + BATCH_SIZE*4096, PROT_READ )) + == NULL ) + goto out; + + region_base = mfn_mapper_base( mapper_handle2 ); + /* Now write out each data page, canonicalising page tables as we go... */ prev_pc = 0; - for ( i = 0; i < srec.nr_pfns; i++ ) + for ( n = 0; n < srec.nr_pfns; ) { - this_pc = (i * 100) / srec.nr_pfns; + this_pc = (n * 100) / srec.nr_pfns; if ( (this_pc - prev_pc) >= 5 ) { verbose_printf("\b\b\b\b%3d%%", this_pc); prev_pc = this_pc; } - mfn = pfn_to_mfn_table[i]; - - ppage = map_pfn_readonly(pm_handle, mfn); - memcpy(page, ppage, PAGE_SIZE); - unmap_pfn(pm_handle, ppage); - - if ( (pfn_type[i] == L1TAB) || (pfn_type[i] == L2TAB) ) - { - for ( j = 0; - j < ((pfn_type[i] == L2TAB) ? - (HYPERVISOR_VIRT_START >> L2_PAGETABLE_SHIFT) : 1024); - j++ ) - { - if ( !(page[j] & _PAGE_PRESENT) ) continue; - mfn = page[j] >> PAGE_SHIFT; - if ( !MFN_IS_IN_PSEUDOPHYS_MAP(mfn) ) - { - ERROR("Frame number in pagetable page is invalid"); - goto out; - } - page[j] &= PAGE_SIZE - 1; - page[j] |= mfn_to_pfn_table[mfn] << PAGE_SHIFT; - } - } - - if ( !checked_write(gfd, page, PAGE_SIZE) ) - { - ERROR("Error when writing to state file"); - goto out; - } + for( j = 0, i = n; j < BATCH_SIZE && i < srec.nr_pfns ; j++, i++ ) + { + pfn_type[j] = live_pfn_to_mfn_table[i]; + } + + + for( j = 0, i = n; j < BATCH_SIZE && i < srec.nr_pfns ; j++, i++ ) + { + /* queue up mappings for all of the pages in this batch */ + +//printf("region n=%d j=%d i=%d mfn=%d\n",n,j,i,live_pfn_to_mfn_table[i]); + mfn_mapper_queue_entry( mapper_handle2, j<<PAGE_SHIFT, + live_pfn_to_mfn_table[i], + PAGE_SIZE ); + } + + if( mfn_mapper_flush_queue(mapper_handle2) ) + { + ERROR("Couldn't map page region"); + goto out; + } + + if ( get_pfn_type_batch(xc_handle, domid, j, pfn_type) ) + { + ERROR("get_pfn_type_batch failed"); + goto out; + } + + for( j = 0, i = n; j < BATCH_SIZE && i < srec.nr_pfns ; j++, i++ ) + { + if((pfn_type[j]>>29) == 7) + { + ERROR("bogus page"); + goto out; + } + + /* canonicalise mfn->pfn */ + pfn_type[j] = (pfn_type[j] & PGT_type_mask) | + live_mfn_to_pfn_table[pfn_type[j]&~PGT_type_mask]; + +/* if(pfn_type[j]>>29) + printf("i=%d type=%d\n",i,pfn_type[i]); */ + } + + + if ( !checked_write(gfd, &j, sizeof(int) ) ) + { + ERROR("Error when writing to state file"); + goto out; + } + + if ( !checked_write(gfd, pfn_type, sizeof(unsigned long)*j ) ) + { + ERROR("Error when writing to state file"); + goto out; + } + + + for( j = 0, i = n; j < BATCH_SIZE && i < srec.nr_pfns ; j++, i++ ) + { + /* write out pages in batch */ + + if ( ((pfn_type[j] & PGT_type_mask) == L1TAB) || + ((pfn_type[j] & PGT_type_mask) == L2TAB) ) + { + + memcpy(page, region_base + (PAGE_SIZE*j), PAGE_SIZE); + + for ( k = 0; + k < (((pfn_type[j] & PGT_type_mask) == L2TAB) ? + (HYPERVISOR_VIRT_START >> L2_PAGETABLE_SHIFT) : 1024); + k++ ) + { + if ( !(page[k] & _PAGE_PRESENT) ) continue; + mfn = page[k] >> PAGE_SHIFT; + if ( !MFN_IS_IN_PSEUDOPHYS_MAP(mfn) ) + { + ERROR("Frame number in pagetable page is invalid"); + goto out; + } + page[k] &= PAGE_SIZE - 1; + page[k] |= live_mfn_to_pfn_table[mfn] << PAGE_SHIFT; + + } + + if ( !checked_write(gfd, page, PAGE_SIZE) ) + { + ERROR("Error when writing to state file"); + goto out; + } + + + } + else + { + if ( !checked_write(gfd, region_base + (PAGE_SIZE*j), PAGE_SIZE) ) + { + ERROR("Error when writing to state file"); + goto out; + } + } + } + + n+=j; /* i is the master loop counter */ } verbose_printf("\b\b\b\b100%%\nMemory saved.\n"); @@ -371,10 +467,19 @@ int xc_linux_save(int xc_handle, /* Success! */ rc = 0; - out: + /* Zero terminate */ + if ( !checked_write(gfd, &rc, sizeof(int)) ) + { + ERROR("Error when writing to state file"); + goto out; + } + + +out: /* Restart the domain if we had to stop it to save its state. */ if ( we_stopped_it ) { + printf("Restart domain\n"); op.cmd = DOM0_STARTDOMAIN; op.u.startdomain.domain = (domid_t)domid; (void)do_dom0_op(xc_handle, &op); @@ -382,13 +487,6 @@ int xc_linux_save(int xc_handle, gzclose(gfd); - if ( pm_handle >= 0 ) - (void)close_pfn_mapper(pm_handle); - - if ( pfn_to_mfn_table != NULL ) - free(pfn_to_mfn_table); - if ( mfn_to_pfn_table != NULL ) - free(mfn_to_pfn_table); if ( pfn_type != NULL ) free(pfn_type); @@ -397,4 +495,6 @@ int xc_linux_save(int xc_handle, unlink(state_file); return !!rc; + + } diff --git a/tools/xc/lib/xc_netbsd_build.c b/tools/xc/lib/xc_netbsd_build.c index cac444bd80..8260c75ea4 100644 --- a/tools/xc/lib/xc_netbsd_build.c +++ b/tools/xc/lib/xc_netbsd_build.c @@ -27,6 +27,7 @@ static long get_tot_pages(int xc_handle, u64 domid) dom0_op_t op; op.cmd = DOM0_GETDOMAININFO; op.u.getdomaininfo.domain = (domid_t)domid; + op.u.getdomaininfo.ctxt = NULL; return (do_dom0_op(xc_handle, &op) < 0) ? -1 : op.u.getdomaininfo.tot_pages; } @@ -59,7 +60,7 @@ static int setup_guestos(int xc_handle, unsigned long tot_pages, unsigned long *virt_startinfo_addr, unsigned long *virt_load_addr, - dom0_builddomain_t *builddomain, + full_execution_context_t *ctxt, const char *cmdline, unsigned long shared_info_frame, unsigned int control_evtchn, @@ -79,8 +80,6 @@ static int setup_guestos(int xc_handle, mmu_t *mmu = NULL; int pm_handle, i; - memset(builddomain, 0, sizeof(*builddomain)); - if ( (pm_handle = init_pfn_mapper((domid_t)dom)) < 0 ) goto error_out; @@ -120,7 +119,7 @@ static int setup_guestos(int xc_handle, */ l2tab = page_array[alloc_index] << PAGE_SHIFT; alloc_index--; - builddomain->ctxt.pt_base = l2tab; + ctxt->pt_base = l2tab; if ( (mmu = init_mmu_updates(xc_handle, dom)) == NULL ) goto error_out; @@ -223,7 +222,7 @@ int xc_netbsd_build(int xc_handle, int kernel_fd = -1; gzFile kernel_gfd = NULL; int rc, i; - full_execution_context_t *ctxt; + full_execution_context_t st_ctxt, *ctxt = &st_ctxt; unsigned long virt_startinfo_addr; if ( (tot_pages = get_tot_pages(xc_handle, domid)) < 0 ) @@ -246,8 +245,15 @@ int xc_netbsd_build(int xc_handle, return 1; } + if ( mlock(&st_ctxt, sizeof(st_ctxt) ) ) + { + PERROR("Unable to mlock ctxt"); + return 1; + } + op.cmd = DOM0_GETDOMAININFO; op.u.getdomaininfo.domain = (domid_t)domid; + op.u.getdomaininfo.ctxt = ctxt; if ( (do_dom0_op(xc_handle, &op) < 0) || ((u64)op.u.getdomaininfo.domain != domid) ) { @@ -255,7 +261,7 @@ int xc_netbsd_build(int xc_handle, goto error_out; } if ( (op.u.getdomaininfo.state != DOMSTATE_STOPPED) || - (op.u.getdomaininfo.ctxt.pt_base != 0) ) + (op.u.getdomaininfo.ctxt->pt_base != 0) ) { ERROR("Domain is already constructed"); goto error_out; @@ -263,7 +269,7 @@ int xc_netbsd_build(int xc_handle, if ( setup_guestos(xc_handle, domid, kernel_gfd, tot_pages, &virt_startinfo_addr, - &load_addr, &launch_op.u.builddomain, cmdline, + &load_addr, &st_ctxt, cmdline, op.u.getdomaininfo.shared_info_frame, control_evtchn, io_priv) < 0 ) { @@ -276,8 +282,6 @@ int xc_netbsd_build(int xc_handle, if( kernel_gfd ) gzclose(kernel_gfd); - ctxt = &launch_op.u.builddomain.ctxt; - ctxt->flags = 0; /* @@ -330,9 +334,11 @@ int xc_netbsd_build(int xc_handle, ctxt->failsafe_callback_cs = FLAT_GUESTOS_CS; ctxt->failsafe_callback_eip = 0; + memset( &launch_op, 0, sizeof(launch_op) ); + launch_op.u.builddomain.domain = (domid_t)domid; launch_op.u.builddomain.num_vifs = 1; - + launch_op.u.builddomain.ctxt = ctxt; launch_op.cmd = DOM0_BUILDDOMAIN; rc = do_dom0_op(xc_handle, &launch_op); diff --git a/tools/xc/lib/xc_private.c b/tools/xc/lib/xc_private.c index 485aa58754..d137176ca8 100644 --- a/tools/xc/lib/xc_private.c +++ b/tools/xc/lib/xc_private.c @@ -45,6 +45,234 @@ void unmap_pfn(int pm_handle, void *vaddr) (void)munmap(vaddr, PAGE_SIZE); } +/*******************/ + +void * mfn_mapper_map_single(int xc_handle, domid_t dom, + int size, int prot, + unsigned long mfn ) +{ + privcmd_mmap_t ioctlx; + privcmd_mmap_entry_t entry; + void *addr; + addr = mmap( NULL, size, prot, MAP_SHARED, xc_handle, 0 ); + if (addr) + { + ioctlx.num=1; + ioctlx.dom=dom; + ioctlx.entry=&entry; + entry.va=(unsigned long) addr; + entry.mfn=mfn; + entry.npages=(size+PAGE_SIZE-1)>>PAGE_SHIFT; + if ( ioctl( xc_handle, IOCTL_PRIVCMD_MMAP, &ioctlx ) <0 ) + return 0; + } + return addr; +} + +mfn_mapper_t * mfn_mapper_init(int xc_handle, domid_t dom, int size, int prot) +{ + mfn_mapper_t * t; + t = calloc( 1, sizeof(mfn_mapper_t)+ + mfn_mapper_queue_size*sizeof(privcmd_mmap_entry_t) ); + if (!t) return NULL; + t->xc_handle = xc_handle; + t->size = size; + t->prot = prot; + t->error = 0; + t->max_queue_size = mfn_mapper_queue_size; + t->addr = mmap( NULL, size, prot, MAP_SHARED, xc_handle, 0 ); + if (!t->addr) + { + free(t); + return NULL; + } + t->ioctl.num = 0; + t->ioctl.dom = dom; + t->ioctl.entry = (privcmd_mmap_entry_t *) &t[1]; + return t; +} + +void * mfn_mapper_base(mfn_mapper_t *t) +{ + return t->addr; +} + +void mfn_mapper_close(mfn_mapper_t *t) +{ + if(t->addr) munmap( t->addr, t->size ); + free(t); +} + +static int __mfn_mapper_flush_queue(mfn_mapper_t *t) +{ + int rc; + rc = ioctl( t->xc_handle, IOCTL_PRIVCMD_MMAP, &t->ioctl ); + t->ioctl.num = 0; + if(rc && !t->error) + t->error = rc; + return rc; +} + +int mfn_mapper_flush_queue(mfn_mapper_t *t) +{ + int rc; + + rc = __mfn_mapper_flush_queue(t); + + if ( t->error ) + { + rc = t->error; + } + + t->error = 0; + return rc; +} + +void * mfn_mapper_queue_entry(mfn_mapper_t *t, int offset, + unsigned long mfn, int size) +{ + privcmd_mmap_entry_t *entry, *prev; + int pages; + + offset &= PAGE_MASK; + pages =(size+PAGE_SIZE-1)>>PAGE_SHIFT; + entry = &t->ioctl.entry[t->ioctl.num]; + + if ( t->ioctl.num > 0 ) + { + prev = &t->ioctl.entry[t->ioctl.num-1]; + + if ( (prev->va+(prev->npages*PAGE_SIZE)) == + ((unsigned long)t->addr+offset) && + (prev->mfn+prev->npages) == mfn ) + { + prev->npages += pages; + return t->addr+offset; + } + } + + entry->va = (unsigned long)t->addr+offset; + entry->mfn = mfn; + entry->npages = pages; + t->ioctl.num++; + + if(t->ioctl.num == t->max_queue_size) + { + if ( __mfn_mapper_flush_queue(t) ) + return 0; + } + + return t->addr+offset; +} + + +/*******************/ + +#if 0 + +mfn_typer_t *mfn_typer_init(int xc_handle, domid_t dom, int num ) +{ + mfn_typer_t *t; + multicall_entry_t *m; + dom0_op_compact_getpageframeinfo_t *d; + + t = calloc(1, sizeof(mfn_typer_t) ); + m = calloc(num, sizeof(multicall_entry_t)); + d = calloc(num, sizeof(dom0_op_compact_getpageframeinfo_t)); + + if (!t || !m || !d) + { + if(t) free(t); + if(m) free(m); + if(d) free(d); + return NULL; + } + +printf("sizeof(m)=%d sizeof(d)=%d m=%p d=%p\n",sizeof(multicall_entry_t), sizeof(dom0_op_compact_getpageframeinfo_t),m,d); + + if ( (mlock(m, sizeof(multicall_entry_t)*num ) != 0) || + (mlock(d, sizeof(dom0_op_compact_getpageframeinfo_t)*num ) != 0) ) + { + PERROR("Could not lock memory for Xen hypercall"); + return NULL; + } + + t->xc_handle = xc_handle; + t->max = num; + t->nr_multicall_ents=0; + t->multicall_list=m; + t->gpf_list=d; + t->dom = dom; + + return t; +} + +void mfn_typer_queue_entry(mfn_typer_t *t, unsigned long mfn ) +{ + int i = t->nr_multicall_ents; + multicall_entry_t *m = &t->multicall_list[i]; + dom0_op_compact_getpageframeinfo_t *d = &t->gpf_list[i]; + + d->cmd = DOM0_GETPAGEFRAMEINFO; + d->interface_version = DOM0_INTERFACE_VERSION; + d->getpageframeinfo.pfn = mfn; + d->getpageframeinfo.domain = t->dom; + d->getpageframeinfo.type = 1000; //~0UL; + + m->op = __HYPERVISOR_dom0_op; + m->args[0] = (unsigned long)d; + + t->nr_multicall_ents++; +} + +int mfn_typer_flush_queue(mfn_typer_t *t) +{ + if (t->nr_multicall_ents == 0) return 0; + do_multicall_op(t->xc_handle, t->multicall_list, t->nr_multicall_ents); + t->nr_multicall_ents = 0; +} + +unsigned int mfn_typer_get_result(mfn_typer_t *t, int idx) +{ + return t->gpf_list[idx].getpageframeinfo.type; +} + +#endif + +/* NB: arr must be mlock'ed */ + +int get_pfn_type_batch(int xc_handle, + u64 dom, int num, unsigned long *arr) +{ + dom0_op_t op; + op.cmd = DOM0_GETPAGEFRAMEINFO2; + op.u.getpageframeinfo2.domain = (domid_t)dom; + op.u.getpageframeinfo2.num = num; + op.u.getpageframeinfo2.array = arr; + return do_dom0_op(xc_handle, &op); +} + +#define GETPFN_ERR (~0U) +unsigned int get_pfn_type(int xc_handle, + unsigned long mfn, + u64 dom) +{ + dom0_op_t op; + op.cmd = DOM0_GETPAGEFRAMEINFO; + op.u.getpageframeinfo.pfn = mfn; + op.u.getpageframeinfo.domain = (domid_t)dom; + if ( do_dom0_op(xc_handle, &op) < 0 ) + { + PERROR("Unexpected failure when getting page frame info!"); + return GETPFN_ERR; + } + return op.u.getpageframeinfo.type; +} + + + +/*******************/ + #define FIRST_MMU_UPDATE 2 static int flush_mmu_updates(int xc_handle, mmu_t *mmu) diff --git a/tools/xc/lib/xc_private.h b/tools/xc/lib/xc_private.h index d4299109e5..eaa301772d 100644 --- a/tools/xc/lib/xc_private.h +++ b/tools/xc/lib/xc_private.h @@ -16,8 +16,6 @@ #include "xc.h" -#include <asm-xen/proc_cmd.h> - /* from xen/include/hypervisor-ifs */ #include <hypervisor-if.h> #include <dom0_ops.h> @@ -25,6 +23,10 @@ #include <event_channel.h> #include <sched_ctl.h> +#include <asm-xen/proc_cmd.h> + + + /* from xend/lib */ #include <domain_controller.h> @@ -108,6 +110,27 @@ static inline int do_dom0_op(int xc_handle, dom0_op_t *op) out1: return ret; } +static inline int do_multicall_op(int xc_handle, + void *call_list, int nr_calls) +{ + int ret = -1; + privcmd_hypercall_t hypercall; + + hypercall.op = __HYPERVISOR_multicall; + hypercall.arg[0] = (unsigned long)call_list; + hypercall.arg[1] = (unsigned long)nr_calls; + + if ( (ret = do_xen_hypercall(xc_handle, &hypercall)) < 0 ) + { + if ( errno == EACCES ) + fprintf(stderr, "Dom0 operation failed -- need to" + " rebuild the user-space tool set?\n"); + goto out1; + } + + out1: return ret; +} + static inline int do_network_op(int xc_handle, network_op_t *op) { int ret = -1; @@ -174,4 +197,77 @@ int add_mmu_update(int xc_handle, mmu_t *mmu, unsigned long ptr, unsigned long val); int finish_mmu_updates(int xc_handle, mmu_t *mmu); + +/* + * ioctl-based mfn mapping interface + */ + +/* +typedef struct privcmd_mmap_entry { + unsigned long va; + unsigned long mfn; + unsigned long npages; +} privcmd_mmap_entry_t; + +typedef struct privcmd_mmap { + int num; + domid_t dom; + privcmd_mmap_entry_t *entry; +} privcmd_mmap_t; +*/ + +#define mfn_mapper_queue_size 128 + +typedef struct mfn_mapper { + int xc_handle; + int size; + int prot; + int error; + int max_queue_size; + void * addr; + privcmd_mmap_t ioctl; + +} mfn_mapper_t; + +void * mfn_mapper_map_single(int xc_handle, domid_t dom, int size, int prot, + unsigned long mfn ); + +mfn_mapper_t * mfn_mapper_init(int xc_handle, domid_t dom, int size, int prot); + +void * mfn_mapper_base(mfn_mapper_t *t); + +void mfn_mapper_close(mfn_mapper_t *t); + +int mfn_mapper_flush_queue(mfn_mapper_t *t); + +void * mfn_mapper_queue_entry(mfn_mapper_t *t, int offset, + unsigned long mfn, int size ); + +/*********************/ + + +#if 0 +typedef struct mfn_typer { + domid_t dom; + int xc_handle; + int max; + dom0_op_t op; +} mfn_typer_t; + + +mfn_typer_t *mfn_typer_init(int xc_handle, domid_t dom, int num ); + +void mfn_typer_queue_entry(mfn_typer_t *t, unsigned long mfn ); + +int mfn_typer_flush_queue(mfn_typer_t *t); +#endif + +int get_pfn_type_batch(int xc_handle, + u64 dom, int num, unsigned long *arr); + +unsigned int get_pfn_type(int xc_handle, + unsigned long mfn, + u64 dom); + + #endif /* __XC_PRIVATE_H__ */ |