* [xl] a fault happens on pv domain reboot @ 2010-06-01 14:42 Sergey Tovpeko 2010-06-01 14:53 ` Gihan Munasinghe 2010-06-01 16:05 ` Stefano Stabellini 0 siblings, 2 replies; 9+ messages in thread From: Sergey Tovpeko @ 2010-06-01 14:42 UTC (permalink / raw) To: xen-devel@lists.xensource.com Hello! After doing reboot from the pv domain, the domain finishes his work. xl destroys it, but then cann't create the new one. It seems a sort of race happens here between domain destroying and creation, but i'm not sure. /var/log/xen/xl-debian-pv.log shows this: Waiting for domain debian-pv (domid 6) to die [pid 5219] Domain 6 is dead Domain 6 needs to be clean: destroying the domain Done. Rebooting now libxl: error: libxl_dom.c:389:userdata_path unable to find domain info for domain 7: Resource temporarily unavailable cannot save config file: Resource temporarily unavailable libxl: error: libxl.c:747:libxl_domain_destroy xc_domain_pause failed for 7 I use the latest xen-unstable.hg. Thank you for your attention. Sergey Tovpeko ^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [xl] a fault happens on pv domain reboot 2010-06-01 14:42 [xl] a fault happens on pv domain reboot Sergey Tovpeko @ 2010-06-01 14:53 ` Gihan Munasinghe 2010-06-01 16:05 ` Stefano Stabellini 1 sibling, 0 replies; 9+ messages in thread From: Gihan Munasinghe @ 2010-06-01 14:53 UTC (permalink / raw) To: Sergey Tovpeko; +Cc: xen-devel@lists.xensource.com Sergey Tovpeko wrote: > Hello! > > After doing reboot from the pv domain, the domain finishes his work. > xl destroys it, but then cann't create the new one. It seems a sort of > race happens here between domain destroying and creation, but i'm not > sure. > I came across with the same problem I put a sleep(2) in the code, I haven't had a chance to fully investigate the the root cause of this See the following patch This is trying to implement xl reboot and xl shutdown. http://lists.xensource.com/archives/html/xen-devel/2010-05/msg01616.html > /var/log/xen/xl-debian-pv.log shows this: > > > Waiting for domain debian-pv (domid 6) to die [pid 5219] > Domain 6 is dead > Domain 6 needs to be clean: destroying the domain > Done. Rebooting now > libxl: error: libxl_dom.c:389:userdata_path unable to find domain info > for domain 7: Resource temporarily unavailable > cannot save config file: Resource temporarily unavailable > libxl: error: libxl.c:747:libxl_domain_destroy xc_domain_pause failed > for 7 > > I use the latest xen-unstable.hg. > > Thank you for your attention. > Sergey Tovpeko > > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.xensource.com > http://lists.xensource.com/xen-devel > > -- Gihan Munasinghe R&D Team Leader Flexiant Ltd. www.flexiant.com ^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [xl] a fault happens on pv domain reboot 2010-06-01 14:42 [xl] a fault happens on pv domain reboot Sergey Tovpeko 2010-06-01 14:53 ` Gihan Munasinghe @ 2010-06-01 16:05 ` Stefano Stabellini 2010-06-01 16:58 ` Ian Jackson 1 sibling, 1 reply; 9+ messages in thread From: Stefano Stabellini @ 2010-06-01 16:05 UTC (permalink / raw) To: Sergey Tovpeko; +Cc: xen-devel@lists.xensource.com, Ian Jackson On Tue, 1 Jun 2010, Sergey Tovpeko wrote: > Hello! > > After doing reboot from the pv domain, the domain finishes his work. > xl destroys it, but then cann't create the new one. It seems a sort of > race happens here between domain destroying and creation, but i'm not sure. > > /var/log/xen/xl-debian-pv.log shows this: > > > Waiting for domain debian-pv (domid 6) to die [pid 5219] > Domain 6 is dead > Domain 6 needs to be clean: destroying the domain > Done. Rebooting now > libxl: error: libxl_dom.c:389:userdata_path unable to find domain info > for domain 7: Resource temporarily unavailable > cannot save config file: Resource temporarily unavailable > libxl: error: libxl.c:747:libxl_domain_destroy xc_domain_pause failed for 7 > > I use the latest xen-unstable.hg. > That means that xc_domain_getinfolist failed with EAGAIN, the reason seems to be that mlock can actually failed with EAGAIN but we don't do anything about it in libxc. If mlock returns EAGAIN we should make sure that the pages are actually available in RAM and try again. ^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [xl] a fault happens on pv domain reboot 2010-06-01 16:05 ` Stefano Stabellini @ 2010-06-01 16:58 ` Ian Jackson 2010-06-02 9:10 ` Sergey Tovpeko 0 siblings, 1 reply; 9+ messages in thread From: Ian Jackson @ 2010-06-01 16:58 UTC (permalink / raw) To: Sergey Tovpeko; +Cc: xen-devel@lists.xensource.com, Stefano Stabellini Stefano Stabellini writes ("Re: [Xen-devel] [xl] a fault happens on pv domain reboot"): > That means that xc_domain_getinfolist failed with EAGAIN, the reason > seems to be that mlock can actually failed with EAGAIN but we don't do > anything about it in libxc. > If mlock returns EAGAIN we should make sure that the pages are actually > available in RAM and try again. So, on that note, could you please try this patch ? Ian. diff -r 267ecb2ee5bf tools/libxc/xc_acm.c --- a/tools/libxc/xc_acm.c Tue Jun 01 14:25:59 2010 +0100 +++ b/tools/libxc/xc_acm.c Tue Jun 01 17:54:51 2010 +0100 @@ -83,7 +83,7 @@ hypercall.op = __HYPERVISOR_xsm_op; hypercall.arg[0] = (unsigned long)&acmctl; - if ( lock_pages(&acmctl, sizeof(acmctl)) != 0) + if ( xc_lock_pages(xch, &acmctl, sizeof(acmctl)) != 0) { PERROR("Could not lock memory for Xen hypercall"); return -EFAULT; diff -r 267ecb2ee5bf tools/libxc/xc_cpupool.c --- a/tools/libxc/xc_cpupool.c Tue Jun 01 14:25:59 2010 +0100 +++ b/tools/libxc/xc_cpupool.c Tue Jun 01 17:54:51 2010 +0100 @@ -71,7 +71,7 @@ set_xen_guest_handle(sysctl.u.cpupool_op.cpumap.bitmap, local); sysctl.u.cpupool_op.cpumap.nr_cpus = sizeof(info->cpumap) * 8; - if ( (err = lock_pages(local, sizeof(local))) != 0 ) + if ( (err = xc_lock_pages(xch, local, sizeof(local))) != 0 ) { PERROR("Could not lock memory for Xen hypercall"); break; @@ -147,7 +147,7 @@ set_xen_guest_handle(sysctl.u.cpupool_op.cpumap.bitmap, local); sysctl.u.cpupool_op.cpumap.nr_cpus = sizeof(*cpumap) * 8; - if ( (err = lock_pages(local, sizeof(local))) != 0 ) + if ( (err = xc_lock_pages(xch, local, sizeof(local))) != 0 ) { PERROR("Could not lock memory for Xen hypercall"); return err; diff -r 267ecb2ee5bf tools/libxc/xc_domain.c --- a/tools/libxc/xc_domain.c Tue Jun 01 14:25:59 2010 +0100 +++ b/tools/libxc/xc_domain.c Tue Jun 01 17:54:51 2010 +0100 @@ -80,7 +80,7 @@ arg.domain_id = domid; arg.reason = reason; - if ( lock_pages(&arg, sizeof(arg)) != 0 ) + if ( xc_lock_pages(xch, &arg, sizeof(arg)) != 0 ) { PERROR("Could not lock memory for Xen hypercall"); goto out1; @@ -119,7 +119,7 @@ domctl.u.vcpuaffinity.cpumap.nr_cpus = cpusize * 8; - if ( lock_pages(local, cpusize) != 0 ) + if ( xc_lock_pages(xch, local, cpusize) != 0 ) { PERROR("Could not lock memory for Xen hypercall"); goto out; @@ -158,7 +158,7 @@ set_xen_guest_handle(domctl.u.vcpuaffinity.cpumap.bitmap, local); domctl.u.vcpuaffinity.cpumap.nr_cpus = cpusize * 8; - if ( lock_pages(local, sizeof(local)) != 0 ) + if ( xc_lock_pages(xch, local, sizeof(local)) != 0 ) { PERROR("Could not lock memory for Xen hypercall"); goto out; @@ -243,7 +243,7 @@ int ret = 0; DECLARE_SYSCTL; - if ( lock_pages(info, max_domains*sizeof(xc_domaininfo_t)) != 0 ) + if ( xc_lock_pages(xch, info, max_domains*sizeof(xc_domaininfo_t)) != 0 ) return -1; sysctl.cmd = XEN_SYSCTL_getdomaininfolist; @@ -276,7 +276,7 @@ set_xen_guest_handle(domctl.u.hvmcontext.buffer, ctxt_buf); if ( ctxt_buf ) - if ( (ret = lock_pages(ctxt_buf, size)) != 0 ) + if ( (ret = xc_lock_pages(xch, ctxt_buf, size)) != 0 ) return ret; ret = do_domctl(xch, &domctl); @@ -308,7 +308,7 @@ domctl.u.hvmcontext_partial.instance = instance; set_xen_guest_handle(domctl.u.hvmcontext_partial.buffer, ctxt_buf); - if ( (ret = lock_pages(ctxt_buf, size)) != 0 ) + if ( (ret = xc_lock_pages(xch, ctxt_buf, size)) != 0 ) return ret; ret = do_domctl(xch, &domctl); @@ -333,7 +333,7 @@ domctl.u.hvmcontext.size = size; set_xen_guest_handle(domctl.u.hvmcontext.buffer, ctxt_buf); - if ( (ret = lock_pages(ctxt_buf, size)) != 0 ) + if ( (ret = xc_lock_pages(xch, ctxt_buf, size)) != 0 ) return ret; ret = do_domctl(xch, &domctl); @@ -358,7 +358,7 @@ set_xen_guest_handle(domctl.u.vcpucontext.ctxt, &ctxt->c); - if ( (rc = lock_pages(ctxt, sz)) != 0 ) + if ( (rc = xc_lock_pages(xch, ctxt, sz)) != 0 ) return rc; rc = do_domctl(xch, &domctl); unlock_pages(ctxt, sz); @@ -446,7 +446,8 @@ set_xen_guest_handle(fmap.map.buffer, &e820); - if ( lock_pages(&fmap, sizeof(fmap)) || lock_pages(&e820, sizeof(e820)) ) + if ( xc_lock_pages(xch, &fmap, sizeof(fmap)) || + xc_lock_pages(xch, &e820, sizeof(e820)) ) { PERROR("Could not lock memory for Xen hypercall"); rc = -1; @@ -522,7 +523,7 @@ domctl.cmd = XEN_DOMCTL_gettscinfo; domctl.domain = (domid_t)domid; set_xen_guest_handle(domctl.u.tsc_info.out_info, &info); - if ( (rc = lock_pages(&info, sizeof(info))) != 0 ) + if ( (rc = xc_lock_pages(xch, &info, sizeof(info))) != 0 ) return rc; rc = do_domctl(xch, &domctl); if ( rc == 0 ) @@ -807,7 +808,7 @@ domctl.u.vcpucontext.vcpu = vcpu; set_xen_guest_handle(domctl.u.vcpucontext.ctxt, &ctxt->c); - if ( (rc = lock_pages(ctxt, sz)) != 0 ) + if ( (rc = xc_lock_pages(xch, ctxt, sz)) != 0 ) return rc; rc = do_domctl(xch, &domctl); @@ -875,7 +876,7 @@ arg.domid = dom; arg.index = param; arg.value = value; - if ( lock_pages(&arg, sizeof(arg)) != 0 ) + if ( xc_lock_pages(handle, &arg, sizeof(arg)) != 0 ) return -1; rc = do_xen_hypercall(handle, &hypercall); unlock_pages(&arg, sizeof(arg)); @@ -893,7 +894,7 @@ hypercall.arg[1] = (unsigned long)&arg; arg.domid = dom; arg.index = param; - if ( lock_pages(&arg, sizeof(arg)) != 0 ) + if ( xc_lock_pages(handle, &arg, sizeof(arg)) != 0 ) return -1; rc = do_xen_hypercall(handle, &hypercall); unlock_pages(&arg, sizeof(arg)); @@ -946,7 +947,7 @@ set_xen_guest_handle(domctl.u.get_device_group.sdev_array, sdev_array); - if ( lock_pages(sdev_array, max_sdevs * sizeof(*sdev_array)) != 0 ) + if ( xc_lock_pages(xch, sdev_array, max_sdevs * sizeof(*sdev_array)) != 0 ) { PERROR("Could not lock memory for xc_get_device_group"); return -ENOMEM; diff -r 267ecb2ee5bf tools/libxc/xc_domain_restore.c --- a/tools/libxc/xc_domain_restore.c Tue Jun 01 14:25:59 2010 +0100 +++ b/tools/libxc/xc_domain_restore.c Tue Jun 01 17:54:51 2010 +0100 @@ -1451,7 +1451,7 @@ memset(region_mfn, 0, ROUNDUP(MAX_BATCH_SIZE * sizeof(xen_pfn_t), PAGE_SHIFT)); - if ( lock_pages(region_mfn, sizeof(xen_pfn_t) * MAX_BATCH_SIZE) ) + if ( xc_lock_pages(xch, region_mfn, sizeof(xen_pfn_t) * MAX_BATCH_SIZE) ) { PERROR("Could not lock region_mfn"); goto out; @@ -1801,7 +1801,7 @@ } } - if ( lock_pages(&ctxt, sizeof(ctxt)) ) + if ( xc_lock_pages(xch, &ctxt, sizeof(ctxt)) ) { PERROR("Unable to lock ctxt"); return 1; diff -r 267ecb2ee5bf tools/libxc/xc_domain_save.c --- a/tools/libxc/xc_domain_save.c Tue Jun 01 14:25:59 2010 +0100 +++ b/tools/libxc/xc_domain_save.c Tue Jun 01 17:54:51 2010 +0100 @@ -1032,14 +1032,14 @@ memset(to_send, 0xff, BITMAP_SIZE); - if ( lock_pages(to_send, BITMAP_SIZE) ) + if ( xc_lock_pages(xch, to_send, BITMAP_SIZE) ) { PERROR("Unable to lock to_send"); return 1; } /* (to fix is local only) */ - if ( lock_pages(to_skip, BITMAP_SIZE) ) + if ( xc_lock_pages(xch, to_skip, BITMAP_SIZE) ) { PERROR("Unable to lock to_skip"); return 1; @@ -1077,7 +1077,7 @@ memset(pfn_type, 0, ROUNDUP(MAX_BATCH_SIZE * sizeof(*pfn_type), PAGE_SHIFT)); - if ( lock_pages(pfn_type, MAX_BATCH_SIZE * sizeof(*pfn_type)) ) + if ( xc_lock_pages(xch, pfn_type, MAX_BATCH_SIZE * sizeof(*pfn_type)) ) { PERROR("Unable to lock pfn_type array"); goto out; diff -r 267ecb2ee5bf tools/libxc/xc_evtchn.c --- a/tools/libxc/xc_evtchn.c Tue Jun 01 14:25:59 2010 +0100 +++ b/tools/libxc/xc_evtchn.c Tue Jun 01 17:54:51 2010 +0100 @@ -19,7 +19,7 @@ hypercall.arg[0] = cmd; hypercall.arg[1] = (unsigned long)arg; - if ( lock_pages(arg, arg_size) != 0 ) + if ( xc_lock_pages(xch, arg, arg_size) != 0 ) { PERROR("do_evtchn_op: arg lock failed"); goto out; diff -r 267ecb2ee5bf tools/libxc/xc_linux.c --- a/tools/libxc/xc_linux.c Tue Jun 01 14:25:59 2010 +0100 +++ b/tools/libxc/xc_linux.c Tue Jun 01 17:54:51 2010 +0100 @@ -724,7 +724,7 @@ hypercall.arg[1] = (unsigned long)op; hypercall.arg[2] = count; - if ( lock_pages(op, count* op_size) != 0 ) + if ( xc_lock_pages(xch, op, count* op_size) != 0 ) { PERROR("Could not lock memory for Xen hypercall"); goto out1; @@ -776,7 +776,7 @@ *gnt_num = query.nr_frames * (PAGE_SIZE / sizeof(grant_entry_v1_t) ); frame_list = malloc(query.nr_frames * sizeof(unsigned long)); - if ( !frame_list || lock_pages(frame_list, + if ( !frame_list || xc_lock_pages(xch, frame_list, query.nr_frames * sizeof(unsigned long)) ) { ERROR("Alloc/lock frame_list in xc_gnttab_map_table\n"); diff -r 267ecb2ee5bf tools/libxc/xc_misc.c --- a/tools/libxc/xc_misc.c Tue Jun 01 14:25:59 2010 +0100 +++ b/tools/libxc/xc_misc.c Tue Jun 01 17:54:51 2010 +0100 @@ -28,7 +28,7 @@ sysctl.u.readconsole.incremental = incremental; } - if ( (ret = lock_pages(buffer, nr_chars)) != 0 ) + if ( (ret = xc_lock_pages(xch, buffer, nr_chars)) != 0 ) return ret; if ( (ret = do_sysctl(xch, &sysctl)) == 0 ) @@ -52,7 +52,7 @@ set_xen_guest_handle(sysctl.u.debug_keys.keys, keys); sysctl.u.debug_keys.nr_keys = len; - if ( (ret = lock_pages(keys, len)) != 0 ) + if ( (ret = xc_lock_pages(xch, keys, len)) != 0 ) return ret; ret = do_sysctl(xch, &sysctl); @@ -140,7 +140,7 @@ DECLARE_HYPERCALL; mc->interface_version = XEN_MCA_INTERFACE_VERSION; - if ( lock_pages(mc, sizeof(mc)) ) + if ( xc_lock_pages(xch, mc, sizeof(mc)) ) { PERROR("Could not lock xen_mc memory"); return -EINVAL; @@ -213,7 +213,7 @@ sysctl.u.getcpuinfo.max_cpus = max_cpus; set_xen_guest_handle(sysctl.u.getcpuinfo.info, info); - if ( (rc = lock_pages(info, max_cpus*sizeof(*info))) != 0 ) + if ( (rc = xc_lock_pages(xch, info, max_cpus*sizeof(*info))) != 0 ) return rc; rc = do_sysctl(xch, &sysctl); @@ -236,7 +236,7 @@ struct xen_hvm_set_pci_intx_level _arg, *arg = &_arg; int rc; - if ( (rc = hcall_buf_prep((void **)&arg, sizeof(*arg))) != 0 ) + if ( (rc = hcall_buf_prep(xch, (void **)&arg, sizeof(*arg))) != 0 ) { PERROR("Could not lock memory"); return rc; @@ -269,7 +269,7 @@ struct xen_hvm_set_isa_irq_level _arg, *arg = &_arg; int rc; - if ( (rc = hcall_buf_prep((void **)&arg, sizeof(*arg))) != 0 ) + if ( (rc = hcall_buf_prep(xch, (void **)&arg, sizeof(*arg))) != 0 ) { PERROR("Could not lock memory"); return rc; @@ -305,7 +305,7 @@ arg.link = link; arg.isa_irq = isa_irq; - if ( (rc = lock_pages(&arg, sizeof(arg))) != 0 ) + if ( (rc = xc_lock_pages(xch, &arg, sizeof(arg))) != 0 ) { PERROR("Could not lock memory"); return rc; @@ -336,7 +336,7 @@ arg.nr = nr; set_xen_guest_handle(arg.dirty_bitmap, (uint8_t *)dirty_bitmap); - if ( (rc = lock_pages(&arg, sizeof(arg))) != 0 ) + if ( (rc = xc_lock_pages(xch, &arg, sizeof(arg))) != 0 ) { PERROR("Could not lock memory"); return rc; @@ -364,7 +364,7 @@ arg.first_pfn = first_pfn; arg.nr = nr; - if ( (rc = lock_pages(&arg, sizeof(arg))) != 0 ) + if ( (rc = xc_lock_pages(xch, &arg, sizeof(arg))) != 0 ) { PERROR("Could not lock memory"); return rc; @@ -393,7 +393,7 @@ arg.first_pfn = first_pfn; arg.nr = nr; - if ( (rc = lock_pages(&arg, sizeof(arg))) != 0 ) + if ( (rc = xc_lock_pages(xch, &arg, sizeof(arg))) != 0 ) { PERROR("Could not lock memory"); return rc; diff -r 267ecb2ee5bf tools/libxc/xc_offline_page.c --- a/tools/libxc/xc_offline_page.c Tue Jun 01 14:25:59 2010 +0100 +++ b/tools/libxc/xc_offline_page.c Tue Jun 01 17:54:51 2010 +0100 @@ -57,7 +57,7 @@ if ( !status || (end < start) ) return -EINVAL; - if (lock_pages(status, sizeof(uint32_t)*(end - start + 1))) + if (xc_lock_pages(xch, status, sizeof(uint32_t)*(end - start + 1))) { ERROR("Could not lock memory for xc_mark_page_online\n"); return -EINVAL; @@ -84,7 +84,7 @@ if ( !status || (end < start) ) return -EINVAL; - if (lock_pages(status, sizeof(uint32_t)*(end - start + 1))) + if (xc_lock_pages(xch, status, sizeof(uint32_t)*(end - start + 1))) { ERROR("Could not lock memory for xc_mark_page_offline"); return -EINVAL; @@ -111,7 +111,7 @@ if ( !status || (end < start) ) return -EINVAL; - if (lock_pages(status, sizeof(uint32_t)*(end - start + 1))) + if (xc_lock_pages(xch, status, sizeof(uint32_t)*(end - start + 1))) { ERROR("Could not lock memory for xc_query_page_offline_status\n"); return -EINVAL; @@ -277,7 +277,7 @@ minfo->pfn_type[i] = pfn_to_mfn(i, minfo->p2m_table, minfo->guest_width); - if ( lock_pages(minfo->pfn_type, minfo->p2m_size * sizeof(*minfo->pfn_type)) ) + if ( xc_lock_pages(xch, minfo->pfn_type, minfo->p2m_size * sizeof(*minfo->pfn_type)) ) { ERROR("Unable to lock pfn_type array"); goto failed; diff -r 267ecb2ee5bf tools/libxc/xc_pm.c --- a/tools/libxc/xc_pm.c Tue Jun 01 14:25:59 2010 +0100 +++ b/tools/libxc/xc_pm.c Tue Jun 01 17:54:51 2010 +0100 @@ -57,11 +57,11 @@ if ( (ret = xc_pm_get_max_px(xch, cpuid, &max_px)) != 0) return ret; - if ( (ret = lock_pages(pxpt->trans_pt, + if ( (ret = xc_lock_pages(xch, pxpt->trans_pt, max_px * max_px * sizeof(uint64_t))) != 0 ) return ret; - if ( (ret = lock_pages(pxpt->pt, + if ( (ret = xc_lock_pages(xch, pxpt->pt, max_px * sizeof(struct xc_px_val))) != 0 ) { unlock_pages(pxpt->trans_pt, max_px * max_px * sizeof(uint64_t)); @@ -132,11 +132,11 @@ if ( (ret = xc_pm_get_max_cx(xch, cpuid, &max_cx)) ) goto unlock_0; - if ( (ret = lock_pages(cxpt, sizeof(struct xc_cx_stat))) ) + if ( (ret = xc_lock_pages(xch, cxpt, sizeof(struct xc_cx_stat))) ) goto unlock_0; - if ( (ret = lock_pages(cxpt->triggers, max_cx * sizeof(uint64_t))) ) + if ( (ret = xc_lock_pages(xch, cxpt->triggers, max_cx * sizeof(uint64_t))) ) goto unlock_1; - if ( (ret = lock_pages(cxpt->residencies, max_cx * sizeof(uint64_t))) ) + if ( (ret = xc_lock_pages(xch, cxpt->residencies, max_cx * sizeof(uint64_t))) ) goto unlock_2; sysctl.cmd = XEN_SYSCTL_get_pmstat; @@ -199,13 +199,13 @@ (!user_para->scaling_available_governors) ) return -EINVAL; - if ( (ret = lock_pages(user_para->affected_cpus, + if ( (ret = xc_lock_pages(xch, user_para->affected_cpus, user_para->cpu_num * sizeof(uint32_t))) ) goto unlock_1; - if ( (ret = lock_pages(user_para->scaling_available_frequencies, + if ( (ret = xc_lock_pages(xch, user_para->scaling_available_frequencies, user_para->freq_num * sizeof(uint32_t))) ) goto unlock_2; - if ( (ret = lock_pages(user_para->scaling_available_governors, + if ( (ret = xc_lock_pages(xch, user_para->scaling_available_governors, user_para->gov_num * CPUFREQ_NAME_LEN * sizeof(char))) ) goto unlock_3; diff -r 267ecb2ee5bf tools/libxc/xc_private.c --- a/tools/libxc/xc_private.c Tue Jun 01 14:25:59 2010 +0100 +++ b/tools/libxc/xc_private.c Tue Jun 01 17:54:51 2010 +0100 @@ -174,7 +174,7 @@ #ifdef __sun__ -int lock_pages(void *addr, size_t len) { return 0; } +int xc_lock_pages(xc_interface *xch, void *addr, size_t len) { return 0; } void unlock_pages(void *addr, size_t len) { } int hcall_buf_prep(void **addr, size_t len) { return 0; } @@ -182,13 +182,40 @@ #else /* !__sun__ */ -int lock_pages(void *addr, size_t len) +int xc_lock_pages(xc_interface *xch, void *addr, size_t len) { int e; void *laddr = (void *)((unsigned long)addr & PAGE_MASK); size_t llen = (len + ((unsigned long)addr - (unsigned long)laddr) + PAGE_SIZE - 1) & PAGE_MASK; - e = mlock(laddr, llen); + size_t offset; + int iterations = 0; + int dummy; + + for (;;) { + e = mlock(laddr, llen); + if (!e) { + if (iterations > 5) + DBGPRINTF("mlock (libxc_lock_pages) (len=%zu)" + " took %d iterations", len, iterations); + return 0; + } + if (errno != EAGAIN) { + PERROR("mlock (libxc_lock_pages) failed (len=%zu)", len); + return e; + } + if (++iterations > 100) { + ERROR("mlock (libxc_lock_pages) too much EAGAIN (len=%zu)", len); + return -1; + } + if (iterations > 10) { + /* max total time: 1/2 * 100 * 2000 us = 1 second */ + usleep(iterations * 2000); + } + for (offset = 0; offset < len; offset += PAGE_SIZE) { + dummy = ((volatile unsigned char*)addr)[offset]; + } + } return e; } @@ -230,7 +257,7 @@ pthread_key_create(&hcall_buf_pkey, _xc_clean_hcall_buf); } -int hcall_buf_prep(void **addr, size_t len) +int hcall_buf_prep(xc_interface *xch, void **addr, size_t len) { struct hcall_buf *hcall_buf; @@ -248,7 +275,7 @@ if ( !hcall_buf->buf ) { hcall_buf->buf = xc_memalign(PAGE_SIZE, PAGE_SIZE); - if ( !hcall_buf->buf || lock_pages(hcall_buf->buf, PAGE_SIZE) ) + if ( !hcall_buf->buf || xc_lock_pages(xch, hcall_buf->buf, PAGE_SIZE) ) { free(hcall_buf->buf); hcall_buf->buf = NULL; @@ -265,7 +292,7 @@ } out: - return lock_pages(*addr, len); + return xc_lock_pages(xch, *addr, len); } void hcall_buf_release(void **addr, size_t len) @@ -307,7 +334,7 @@ DECLARE_HYPERCALL; long ret = -EINVAL; - if ( hcall_buf_prep((void **)&op, nr_ops*sizeof(*op)) != 0 ) + if ( hcall_buf_prep(xch, (void **)&op, nr_ops*sizeof(*op)) != 0 ) { PERROR("Could not lock memory for Xen hypercall"); goto out1; @@ -341,7 +368,7 @@ hypercall.arg[2] = 0; hypercall.arg[3] = mmu->subject; - if ( lock_pages(mmu->updates, sizeof(mmu->updates)) != 0 ) + if ( xc_lock_pages(xch, mmu->updates, sizeof(mmu->updates)) != 0 ) { PERROR("flush_mmu_updates: mmu updates lock_pages failed"); err = 1; @@ -408,14 +435,14 @@ case XENMEM_increase_reservation: case XENMEM_decrease_reservation: case XENMEM_populate_physmap: - if ( lock_pages(reservation, sizeof(*reservation)) != 0 ) + if ( xc_lock_pages(xch, reservation, sizeof(*reservation)) != 0 ) { PERROR("Could not lock"); goto out1; } get_xen_guest_handle(extent_start, reservation->extent_start); if ( (extent_start != NULL) && - (lock_pages(extent_start, + (xc_lock_pages(xch, extent_start, reservation->nr_extents * sizeof(xen_pfn_t)) != 0) ) { PERROR("Could not lock"); @@ -424,13 +451,13 @@ } break; case XENMEM_machphys_mfn_list: - if ( lock_pages(xmml, sizeof(*xmml)) != 0 ) + if ( xc_lock_pages(xch, xmml, sizeof(*xmml)) != 0 ) { PERROR("Could not lock"); goto out1; } get_xen_guest_handle(extent_start, xmml->extent_start); - if ( lock_pages(extent_start, + if ( xc_lock_pages(xch, extent_start, xmml->max_extents * sizeof(xen_pfn_t)) != 0 ) { PERROR("Could not lock"); @@ -439,7 +466,7 @@ } break; case XENMEM_add_to_physmap: - if ( lock_pages(arg, sizeof(struct xen_add_to_physmap)) ) + if ( xc_lock_pages(xch, arg, sizeof(struct xen_add_to_physmap)) ) { PERROR("Could not lock"); goto out1; @@ -448,7 +475,7 @@ case XENMEM_current_reservation: case XENMEM_maximum_reservation: case XENMEM_maximum_gpfn: - if ( lock_pages(arg, sizeof(domid_t)) ) + if ( xc_lock_pages(xch, arg, sizeof(domid_t)) ) { PERROR("Could not lock"); goto out1; @@ -456,7 +483,7 @@ break; case XENMEM_set_pod_target: case XENMEM_get_pod_target: - if ( lock_pages(arg, sizeof(struct xen_pod_target)) ) + if ( xc_lock_pages(xch, arg, sizeof(struct xen_pod_target)) ) { PERROR("Could not lock"); goto out1; @@ -535,7 +562,7 @@ memset(pfn_buf, 0, max_pfns * sizeof(*pfn_buf)); #endif - if ( lock_pages(pfn_buf, max_pfns * sizeof(*pfn_buf)) != 0 ) + if ( xc_lock_pages(xch, pfn_buf, max_pfns * sizeof(*pfn_buf)) != 0 ) { PERROR("xc_get_pfn_list: pfn_buf lock failed"); return -1; @@ -618,7 +645,7 @@ break; } - if ( (argsize != 0) && (lock_pages(arg, argsize) != 0) ) + if ( (argsize != 0) && (xc_lock_pages(xch, arg, argsize) != 0) ) { PERROR("Could not lock memory for version hypercall"); return -ENOMEM; diff -r 267ecb2ee5bf tools/libxc/xc_private.h --- a/tools/libxc/xc_private.h Tue Jun 01 14:25:59 2010 +0100 +++ b/tools/libxc/xc_private.h Tue Jun 01 17:54:51 2010 +0100 @@ -85,10 +85,10 @@ void *xc_memalign(size_t alignment, size_t size); -int lock_pages(void *addr, size_t len); +int xc_lock_pages(xc_interface*, void *addr, size_t len); void unlock_pages(void *addr, size_t len); -int hcall_buf_prep(void **addr, size_t len); +int hcall_buf_prep(xc_interface*, void **addr, size_t len); void hcall_buf_release(void **addr, size_t len); static inline void safe_munlock(const void *addr, size_t len) @@ -117,7 +117,7 @@ DECLARE_HYPERCALL; - if ( hcall_buf_prep(&op, len) != 0 ) + if ( hcall_buf_prep(xch, &op, len) != 0 ) { PERROR("Could not lock memory for Xen hypercall"); goto out1; @@ -145,7 +145,7 @@ int ret = -1; DECLARE_HYPERCALL; - if ( hcall_buf_prep((void **)&domctl, sizeof(*domctl)) != 0 ) + if ( hcall_buf_prep(xch, (void **)&domctl, sizeof(*domctl)) != 0 ) { PERROR("Could not lock memory for Xen hypercall"); goto out1; @@ -174,7 +174,7 @@ int ret = -1; DECLARE_HYPERCALL; - if ( hcall_buf_prep((void **)&sysctl, sizeof(*sysctl)) != 0 ) + if ( hcall_buf_prep(xch, (void **)&sysctl, sizeof(*sysctl)) != 0 ) { PERROR("Could not lock memory for Xen hypercall"); goto out1; diff -r 267ecb2ee5bf tools/libxc/xc_resume.c --- a/tools/libxc/xc_resume.c Tue Jun 01 14:25:59 2010 +0100 +++ b/tools/libxc/xc_resume.c Tue Jun 01 17:54:51 2010 +0100 @@ -180,7 +180,7 @@ goto out; } - if ( lock_pages(&ctxt, sizeof(ctxt)) ) + if ( xc_lock_pages(xch, &ctxt, sizeof(ctxt)) ) { ERROR("Unable to lock ctxt"); goto out; diff -r 267ecb2ee5bf tools/libxc/xc_tbuf.c --- a/tools/libxc/xc_tbuf.c Tue Jun 01 14:25:59 2010 +0100 +++ b/tools/libxc/xc_tbuf.c Tue Jun 01 17:54:51 2010 +0100 @@ -120,7 +120,7 @@ set_xen_guest_handle(sysctl.u.tbuf_op.cpu_mask.bitmap, bytemap); sysctl.u.tbuf_op.cpu_mask.nr_cpus = sizeof(bytemap) * 8; - if ( lock_pages(&bytemap, sizeof(bytemap)) != 0 ) + if ( xc_lock_pages(xch, &bytemap, sizeof(bytemap)) != 0 ) { PERROR("Could not lock memory for Xen hypercall"); goto out; diff -r 267ecb2ee5bf tools/libxc/xc_tmem.c --- a/tools/libxc/xc_tmem.c Tue Jun 01 14:25:59 2010 +0100 +++ b/tools/libxc/xc_tmem.c Tue Jun 01 17:54:51 2010 +0100 @@ -14,7 +14,7 @@ hypercall.op = __HYPERVISOR_tmem_op; hypercall.arg[0] = (unsigned long)op; - if (lock_pages(op, sizeof(*op)) != 0) + if (xc_lock_pages(xch, op, sizeof(*op)) != 0) { PERROR("Could not lock memory for Xen hypercall"); return -EFAULT; @@ -52,7 +52,7 @@ op.u.ctrl.arg3 = arg3; if (subop == TMEMC_LIST) { - if ((arg1 != 0) && (lock_pages(buf, arg1) != 0)) + if ((arg1 != 0) && (xc_lock_pages(xch, buf, arg1) != 0)) { PERROR("Could not lock memory for Xen hypercall"); return -ENOMEM; ^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [xl] a fault happens on pv domain reboot 2010-06-01 16:58 ` Ian Jackson @ 2010-06-02 9:10 ` Sergey Tovpeko 2010-06-02 9:27 ` Keir Fraser 2010-06-02 10:11 ` Ian Jackson 0 siblings, 2 replies; 9+ messages in thread From: Sergey Tovpeko @ 2010-06-02 9:10 UTC (permalink / raw) To: Ian Jackson; +Cc: xen-devel@lists.xensource.com, Stefano Stabellini The mlock issue isn't my case. I mean the patch didn't solve the problem. The mentioned mlock returned success with the first iteration. I thought below patch (trick) solves the issue, but it only decreases the probability of the bug. diff -r 458589ad3793 tools/libxl/xl_cmdimpl.c --- a/tools/libxl/xl_cmdimpl.c Tue Jun 01 07:06:50 2010 +0100 +++ b/tools/libxl/xl_cmdimpl.c Wed Jun 02 16:57:59 2010 +0400 @@ -1159,6 +1159,15 @@ libxl_free_waiter(w2); free(w1); free(w2); + while (1) { + struct libxl_dominfo info_t; + if (libxl_domain_info(&ctx, &info_t, domid)) + break; + sleep(1); + } LOG("Done. Rebooting now"); goto start; } > Stefano Stabellini writes ("Re: [Xen-devel] [xl] a fault happens on pv domain reboot"): > >> That means that xc_domain_getinfolist failed with EAGAIN, the reason >> seems to be that mlock can actually failed with EAGAIN but we don't do >> anything about it in libxc. >> If mlock returns EAGAIN we should make sure that the pages are actually >> available in RAM and try again. >> > > So, on that note, could you please try this patch ? > > Ian. > > diff -r 267ecb2ee5bf tools/libxc/xc_acm.c > --- a/tools/libxc/xc_acm.c Tue Jun 01 14:25:59 2010 +0100 > +++ b/tools/libxc/xc_acm.c Tue Jun 01 17:54:51 2010 +0100 > @@ -83,7 +83,7 @@ > > hypercall.op = __HYPERVISOR_xsm_op; > hypercall.arg[0] = (unsigned long)&acmctl; > - if ( lock_pages(&acmctl, sizeof(acmctl)) != 0) > + if ( xc_lock_pages(xch, &acmctl, sizeof(acmctl)) != 0) > { > PERROR("Could not lock memory for Xen hypercall"); > return -EFAULT; > diff -r 267ecb2ee5bf tools/libxc/xc_cpupool.c > --- a/tools/libxc/xc_cpupool.c Tue Jun 01 14:25:59 2010 +0100 > +++ b/tools/libxc/xc_cpupool.c Tue Jun 01 17:54:51 2010 +0100 > @@ -71,7 +71,7 @@ > set_xen_guest_handle(sysctl.u.cpupool_op.cpumap.bitmap, local); > sysctl.u.cpupool_op.cpumap.nr_cpus = sizeof(info->cpumap) * 8; > > - if ( (err = lock_pages(local, sizeof(local))) != 0 ) > + if ( (err = xc_lock_pages(xch, local, sizeof(local))) != 0 ) > { > PERROR("Could not lock memory for Xen hypercall"); > break; > @@ -147,7 +147,7 @@ > set_xen_guest_handle(sysctl.u.cpupool_op.cpumap.bitmap, local); > sysctl.u.cpupool_op.cpumap.nr_cpus = sizeof(*cpumap) * 8; > > - if ( (err = lock_pages(local, sizeof(local))) != 0 ) > + if ( (err = xc_lock_pages(xch, local, sizeof(local))) != 0 ) > { > PERROR("Could not lock memory for Xen hypercall"); > return err; > diff -r 267ecb2ee5bf tools/libxc/xc_domain.c > --- a/tools/libxc/xc_domain.c Tue Jun 01 14:25:59 2010 +0100 > +++ b/tools/libxc/xc_domain.c Tue Jun 01 17:54:51 2010 +0100 > @@ -80,7 +80,7 @@ > arg.domain_id = domid; > arg.reason = reason; > > - if ( lock_pages(&arg, sizeof(arg)) != 0 ) > + if ( xc_lock_pages(xch, &arg, sizeof(arg)) != 0 ) > { > PERROR("Could not lock memory for Xen hypercall"); > goto out1; > @@ -119,7 +119,7 @@ > > domctl.u.vcpuaffinity.cpumap.nr_cpus = cpusize * 8; > > - if ( lock_pages(local, cpusize) != 0 ) > + if ( xc_lock_pages(xch, local, cpusize) != 0 ) > { > PERROR("Could not lock memory for Xen hypercall"); > goto out; > @@ -158,7 +158,7 @@ > set_xen_guest_handle(domctl.u.vcpuaffinity.cpumap.bitmap, local); > domctl.u.vcpuaffinity.cpumap.nr_cpus = cpusize * 8; > > - if ( lock_pages(local, sizeof(local)) != 0 ) > + if ( xc_lock_pages(xch, local, sizeof(local)) != 0 ) > { > PERROR("Could not lock memory for Xen hypercall"); > goto out; > @@ -243,7 +243,7 @@ > int ret = 0; > DECLARE_SYSCTL; > > - if ( lock_pages(info, max_domains*sizeof(xc_domaininfo_t)) != 0 ) > + if ( xc_lock_pages(xch, info, max_domains*sizeof(xc_domaininfo_t)) != 0 ) > return -1; > > sysctl.cmd = XEN_SYSCTL_getdomaininfolist; > @@ -276,7 +276,7 @@ > set_xen_guest_handle(domctl.u.hvmcontext.buffer, ctxt_buf); > > if ( ctxt_buf ) > - if ( (ret = lock_pages(ctxt_buf, size)) != 0 ) > + if ( (ret = xc_lock_pages(xch, ctxt_buf, size)) != 0 ) > return ret; > > ret = do_domctl(xch, &domctl); > @@ -308,7 +308,7 @@ > domctl.u.hvmcontext_partial.instance = instance; > set_xen_guest_handle(domctl.u.hvmcontext_partial.buffer, ctxt_buf); > > - if ( (ret = lock_pages(ctxt_buf, size)) != 0 ) > + if ( (ret = xc_lock_pages(xch, ctxt_buf, size)) != 0 ) > return ret; > > ret = do_domctl(xch, &domctl); > @@ -333,7 +333,7 @@ > domctl.u.hvmcontext.size = size; > set_xen_guest_handle(domctl.u.hvmcontext.buffer, ctxt_buf); > > - if ( (ret = lock_pages(ctxt_buf, size)) != 0 ) > + if ( (ret = xc_lock_pages(xch, ctxt_buf, size)) != 0 ) > return ret; > > ret = do_domctl(xch, &domctl); > @@ -358,7 +358,7 @@ > set_xen_guest_handle(domctl.u.vcpucontext.ctxt, &ctxt->c); > > > - if ( (rc = lock_pages(ctxt, sz)) != 0 ) > + if ( (rc = xc_lock_pages(xch, ctxt, sz)) != 0 ) > return rc; > rc = do_domctl(xch, &domctl); > unlock_pages(ctxt, sz); > @@ -446,7 +446,8 @@ > > set_xen_guest_handle(fmap.map.buffer, &e820); > > - if ( lock_pages(&fmap, sizeof(fmap)) || lock_pages(&e820, sizeof(e820)) ) > + if ( xc_lock_pages(xch, &fmap, sizeof(fmap)) || > + xc_lock_pages(xch, &e820, sizeof(e820)) ) > { > PERROR("Could not lock memory for Xen hypercall"); > rc = -1; > @@ -522,7 +523,7 @@ > domctl.cmd = XEN_DOMCTL_gettscinfo; > domctl.domain = (domid_t)domid; > set_xen_guest_handle(domctl.u.tsc_info.out_info, &info); > - if ( (rc = lock_pages(&info, sizeof(info))) != 0 ) > + if ( (rc = xc_lock_pages(xch, &info, sizeof(info))) != 0 ) > return rc; > rc = do_domctl(xch, &domctl); > if ( rc == 0 ) > @@ -807,7 +808,7 @@ > domctl.u.vcpucontext.vcpu = vcpu; > set_xen_guest_handle(domctl.u.vcpucontext.ctxt, &ctxt->c); > > - if ( (rc = lock_pages(ctxt, sz)) != 0 ) > + if ( (rc = xc_lock_pages(xch, ctxt, sz)) != 0 ) > return rc; > rc = do_domctl(xch, &domctl); > > @@ -875,7 +876,7 @@ > arg.domid = dom; > arg.index = param; > arg.value = value; > - if ( lock_pages(&arg, sizeof(arg)) != 0 ) > + if ( xc_lock_pages(handle, &arg, sizeof(arg)) != 0 ) > return -1; > rc = do_xen_hypercall(handle, &hypercall); > unlock_pages(&arg, sizeof(arg)); > @@ -893,7 +894,7 @@ > hypercall.arg[1] = (unsigned long)&arg; > arg.domid = dom; > arg.index = param; > - if ( lock_pages(&arg, sizeof(arg)) != 0 ) > + if ( xc_lock_pages(handle, &arg, sizeof(arg)) != 0 ) > return -1; > rc = do_xen_hypercall(handle, &hypercall); > unlock_pages(&arg, sizeof(arg)); > @@ -946,7 +947,7 @@ > > set_xen_guest_handle(domctl.u.get_device_group.sdev_array, sdev_array); > > - if ( lock_pages(sdev_array, max_sdevs * sizeof(*sdev_array)) != 0 ) > + if ( xc_lock_pages(xch, sdev_array, max_sdevs * sizeof(*sdev_array)) != 0 ) > { > PERROR("Could not lock memory for xc_get_device_group"); > return -ENOMEM; > diff -r 267ecb2ee5bf tools/libxc/xc_domain_restore.c > --- a/tools/libxc/xc_domain_restore.c Tue Jun 01 14:25:59 2010 +0100 > +++ b/tools/libxc/xc_domain_restore.c Tue Jun 01 17:54:51 2010 +0100 > @@ -1451,7 +1451,7 @@ > memset(region_mfn, 0, > ROUNDUP(MAX_BATCH_SIZE * sizeof(xen_pfn_t), PAGE_SHIFT)); > > - if ( lock_pages(region_mfn, sizeof(xen_pfn_t) * MAX_BATCH_SIZE) ) > + if ( xc_lock_pages(xch, region_mfn, sizeof(xen_pfn_t) * MAX_BATCH_SIZE) ) > { > PERROR("Could not lock region_mfn"); > goto out; > @@ -1801,7 +1801,7 @@ > } > } > > - if ( lock_pages(&ctxt, sizeof(ctxt)) ) > + if ( xc_lock_pages(xch, &ctxt, sizeof(ctxt)) ) > { > PERROR("Unable to lock ctxt"); > return 1; > diff -r 267ecb2ee5bf tools/libxc/xc_domain_save.c > --- a/tools/libxc/xc_domain_save.c Tue Jun 01 14:25:59 2010 +0100 > +++ b/tools/libxc/xc_domain_save.c Tue Jun 01 17:54:51 2010 +0100 > @@ -1032,14 +1032,14 @@ > > memset(to_send, 0xff, BITMAP_SIZE); > > - if ( lock_pages(to_send, BITMAP_SIZE) ) > + if ( xc_lock_pages(xch, to_send, BITMAP_SIZE) ) > { > PERROR("Unable to lock to_send"); > return 1; > } > > /* (to fix is local only) */ > - if ( lock_pages(to_skip, BITMAP_SIZE) ) > + if ( xc_lock_pages(xch, to_skip, BITMAP_SIZE) ) > { > PERROR("Unable to lock to_skip"); > return 1; > @@ -1077,7 +1077,7 @@ > memset(pfn_type, 0, > ROUNDUP(MAX_BATCH_SIZE * sizeof(*pfn_type), PAGE_SHIFT)); > > - if ( lock_pages(pfn_type, MAX_BATCH_SIZE * sizeof(*pfn_type)) ) > + if ( xc_lock_pages(xch, pfn_type, MAX_BATCH_SIZE * sizeof(*pfn_type)) ) > { > PERROR("Unable to lock pfn_type array"); > goto out; > diff -r 267ecb2ee5bf tools/libxc/xc_evtchn.c > --- a/tools/libxc/xc_evtchn.c Tue Jun 01 14:25:59 2010 +0100 > +++ b/tools/libxc/xc_evtchn.c Tue Jun 01 17:54:51 2010 +0100 > @@ -19,7 +19,7 @@ > hypercall.arg[0] = cmd; > hypercall.arg[1] = (unsigned long)arg; > > - if ( lock_pages(arg, arg_size) != 0 ) > + if ( xc_lock_pages(xch, arg, arg_size) != 0 ) > { > PERROR("do_evtchn_op: arg lock failed"); > goto out; > diff -r 267ecb2ee5bf tools/libxc/xc_linux.c > --- a/tools/libxc/xc_linux.c Tue Jun 01 14:25:59 2010 +0100 > +++ b/tools/libxc/xc_linux.c Tue Jun 01 17:54:51 2010 +0100 > @@ -724,7 +724,7 @@ > hypercall.arg[1] = (unsigned long)op; > hypercall.arg[2] = count; > > - if ( lock_pages(op, count* op_size) != 0 ) > + if ( xc_lock_pages(xch, op, count* op_size) != 0 ) > { > PERROR("Could not lock memory for Xen hypercall"); > goto out1; > @@ -776,7 +776,7 @@ > *gnt_num = query.nr_frames * (PAGE_SIZE / sizeof(grant_entry_v1_t) ); > > frame_list = malloc(query.nr_frames * sizeof(unsigned long)); > - if ( !frame_list || lock_pages(frame_list, > + if ( !frame_list || xc_lock_pages(xch, frame_list, > query.nr_frames * sizeof(unsigned long)) ) > { > ERROR("Alloc/lock frame_list in xc_gnttab_map_table\n"); > diff -r 267ecb2ee5bf tools/libxc/xc_misc.c > --- a/tools/libxc/xc_misc.c Tue Jun 01 14:25:59 2010 +0100 > +++ b/tools/libxc/xc_misc.c Tue Jun 01 17:54:51 2010 +0100 > @@ -28,7 +28,7 @@ > sysctl.u.readconsole.incremental = incremental; > } > > - if ( (ret = lock_pages(buffer, nr_chars)) != 0 ) > + if ( (ret = xc_lock_pages(xch, buffer, nr_chars)) != 0 ) > return ret; > > if ( (ret = do_sysctl(xch, &sysctl)) == 0 ) > @@ -52,7 +52,7 @@ > set_xen_guest_handle(sysctl.u.debug_keys.keys, keys); > sysctl.u.debug_keys.nr_keys = len; > > - if ( (ret = lock_pages(keys, len)) != 0 ) > + if ( (ret = xc_lock_pages(xch, keys, len)) != 0 ) > return ret; > > ret = do_sysctl(xch, &sysctl); > @@ -140,7 +140,7 @@ > DECLARE_HYPERCALL; > > mc->interface_version = XEN_MCA_INTERFACE_VERSION; > - if ( lock_pages(mc, sizeof(mc)) ) > + if ( xc_lock_pages(xch, mc, sizeof(mc)) ) > { > PERROR("Could not lock xen_mc memory"); > return -EINVAL; > @@ -213,7 +213,7 @@ > sysctl.u.getcpuinfo.max_cpus = max_cpus; > set_xen_guest_handle(sysctl.u.getcpuinfo.info, info); > > - if ( (rc = lock_pages(info, max_cpus*sizeof(*info))) != 0 ) > + if ( (rc = xc_lock_pages(xch, info, max_cpus*sizeof(*info))) != 0 ) > return rc; > > rc = do_sysctl(xch, &sysctl); > @@ -236,7 +236,7 @@ > struct xen_hvm_set_pci_intx_level _arg, *arg = &_arg; > int rc; > > - if ( (rc = hcall_buf_prep((void **)&arg, sizeof(*arg))) != 0 ) > + if ( (rc = hcall_buf_prep(xch, (void **)&arg, sizeof(*arg))) != 0 ) > { > PERROR("Could not lock memory"); > return rc; > @@ -269,7 +269,7 @@ > struct xen_hvm_set_isa_irq_level _arg, *arg = &_arg; > int rc; > > - if ( (rc = hcall_buf_prep((void **)&arg, sizeof(*arg))) != 0 ) > + if ( (rc = hcall_buf_prep(xch, (void **)&arg, sizeof(*arg))) != 0 ) > { > PERROR("Could not lock memory"); > return rc; > @@ -305,7 +305,7 @@ > arg.link = link; > arg.isa_irq = isa_irq; > > - if ( (rc = lock_pages(&arg, sizeof(arg))) != 0 ) > + if ( (rc = xc_lock_pages(xch, &arg, sizeof(arg))) != 0 ) > { > PERROR("Could not lock memory"); > return rc; > @@ -336,7 +336,7 @@ > arg.nr = nr; > set_xen_guest_handle(arg.dirty_bitmap, (uint8_t *)dirty_bitmap); > > - if ( (rc = lock_pages(&arg, sizeof(arg))) != 0 ) > + if ( (rc = xc_lock_pages(xch, &arg, sizeof(arg))) != 0 ) > { > PERROR("Could not lock memory"); > return rc; > @@ -364,7 +364,7 @@ > arg.first_pfn = first_pfn; > arg.nr = nr; > > - if ( (rc = lock_pages(&arg, sizeof(arg))) != 0 ) > + if ( (rc = xc_lock_pages(xch, &arg, sizeof(arg))) != 0 ) > { > PERROR("Could not lock memory"); > return rc; > @@ -393,7 +393,7 @@ > arg.first_pfn = first_pfn; > arg.nr = nr; > > - if ( (rc = lock_pages(&arg, sizeof(arg))) != 0 ) > + if ( (rc = xc_lock_pages(xch, &arg, sizeof(arg))) != 0 ) > { > PERROR("Could not lock memory"); > return rc; > diff -r 267ecb2ee5bf tools/libxc/xc_offline_page.c > --- a/tools/libxc/xc_offline_page.c Tue Jun 01 14:25:59 2010 +0100 > +++ b/tools/libxc/xc_offline_page.c Tue Jun 01 17:54:51 2010 +0100 > @@ -57,7 +57,7 @@ > if ( !status || (end < start) ) > return -EINVAL; > > - if (lock_pages(status, sizeof(uint32_t)*(end - start + 1))) > + if (xc_lock_pages(xch, status, sizeof(uint32_t)*(end - start + 1))) > { > ERROR("Could not lock memory for xc_mark_page_online\n"); > return -EINVAL; > @@ -84,7 +84,7 @@ > if ( !status || (end < start) ) > return -EINVAL; > > - if (lock_pages(status, sizeof(uint32_t)*(end - start + 1))) > + if (xc_lock_pages(xch, status, sizeof(uint32_t)*(end - start + 1))) > { > ERROR("Could not lock memory for xc_mark_page_offline"); > return -EINVAL; > @@ -111,7 +111,7 @@ > if ( !status || (end < start) ) > return -EINVAL; > > - if (lock_pages(status, sizeof(uint32_t)*(end - start + 1))) > + if (xc_lock_pages(xch, status, sizeof(uint32_t)*(end - start + 1))) > { > ERROR("Could not lock memory for xc_query_page_offline_status\n"); > return -EINVAL; > @@ -277,7 +277,7 @@ > minfo->pfn_type[i] = pfn_to_mfn(i, minfo->p2m_table, > minfo->guest_width); > > - if ( lock_pages(minfo->pfn_type, minfo->p2m_size * sizeof(*minfo->pfn_type)) ) > + if ( xc_lock_pages(xch, minfo->pfn_type, minfo->p2m_size * sizeof(*minfo->pfn_type)) ) > { > ERROR("Unable to lock pfn_type array"); > goto failed; > diff -r 267ecb2ee5bf tools/libxc/xc_pm.c > --- a/tools/libxc/xc_pm.c Tue Jun 01 14:25:59 2010 +0100 > +++ b/tools/libxc/xc_pm.c Tue Jun 01 17:54:51 2010 +0100 > @@ -57,11 +57,11 @@ > if ( (ret = xc_pm_get_max_px(xch, cpuid, &max_px)) != 0) > return ret; > > - if ( (ret = lock_pages(pxpt->trans_pt, > + if ( (ret = xc_lock_pages(xch, pxpt->trans_pt, > max_px * max_px * sizeof(uint64_t))) != 0 ) > return ret; > > - if ( (ret = lock_pages(pxpt->pt, > + if ( (ret = xc_lock_pages(xch, pxpt->pt, > max_px * sizeof(struct xc_px_val))) != 0 ) > { > unlock_pages(pxpt->trans_pt, max_px * max_px * sizeof(uint64_t)); > @@ -132,11 +132,11 @@ > if ( (ret = xc_pm_get_max_cx(xch, cpuid, &max_cx)) ) > goto unlock_0; > > - if ( (ret = lock_pages(cxpt, sizeof(struct xc_cx_stat))) ) > + if ( (ret = xc_lock_pages(xch, cxpt, sizeof(struct xc_cx_stat))) ) > goto unlock_0; > - if ( (ret = lock_pages(cxpt->triggers, max_cx * sizeof(uint64_t))) ) > + if ( (ret = xc_lock_pages(xch, cxpt->triggers, max_cx * sizeof(uint64_t))) ) > goto unlock_1; > - if ( (ret = lock_pages(cxpt->residencies, max_cx * sizeof(uint64_t))) ) > + if ( (ret = xc_lock_pages(xch, cxpt->residencies, max_cx * sizeof(uint64_t))) ) > goto unlock_2; > > sysctl.cmd = XEN_SYSCTL_get_pmstat; > @@ -199,13 +199,13 @@ > (!user_para->scaling_available_governors) ) > return -EINVAL; > > - if ( (ret = lock_pages(user_para->affected_cpus, > + if ( (ret = xc_lock_pages(xch, user_para->affected_cpus, > user_para->cpu_num * sizeof(uint32_t))) ) > goto unlock_1; > - if ( (ret = lock_pages(user_para->scaling_available_frequencies, > + if ( (ret = xc_lock_pages(xch, user_para->scaling_available_frequencies, > user_para->freq_num * sizeof(uint32_t))) ) > goto unlock_2; > - if ( (ret = lock_pages(user_para->scaling_available_governors, > + if ( (ret = xc_lock_pages(xch, user_para->scaling_available_governors, > user_para->gov_num * CPUFREQ_NAME_LEN * sizeof(char))) ) > goto unlock_3; > > diff -r 267ecb2ee5bf tools/libxc/xc_private.c > --- a/tools/libxc/xc_private.c Tue Jun 01 14:25:59 2010 +0100 > +++ b/tools/libxc/xc_private.c Tue Jun 01 17:54:51 2010 +0100 > @@ -174,7 +174,7 @@ > > #ifdef __sun__ > > -int lock_pages(void *addr, size_t len) { return 0; } > +int xc_lock_pages(xc_interface *xch, void *addr, size_t len) { return 0; } > void unlock_pages(void *addr, size_t len) { } > > int hcall_buf_prep(void **addr, size_t len) { return 0; } > @@ -182,13 +182,40 @@ > > #else /* !__sun__ */ > > -int lock_pages(void *addr, size_t len) > +int xc_lock_pages(xc_interface *xch, void *addr, size_t len) > { > int e; > void *laddr = (void *)((unsigned long)addr & PAGE_MASK); > size_t llen = (len + ((unsigned long)addr - (unsigned long)laddr) + > PAGE_SIZE - 1) & PAGE_MASK; > - e = mlock(laddr, llen); > + size_t offset; > + int iterations = 0; > + int dummy; > + > + for (;;) { > + e = mlock(laddr, llen); > + if (!e) { > + if (iterations > 5) > + DBGPRINTF("mlock (libxc_lock_pages) (len=%zu)" > + " took %d iterations", len, iterations); > + return 0; > + } > + if (errno != EAGAIN) { > + PERROR("mlock (libxc_lock_pages) failed (len=%zu)", len); > + return e; > + } > + if (++iterations > 100) { > + ERROR("mlock (libxc_lock_pages) too much EAGAIN (len=%zu)", len); > + return -1; > + } > + if (iterations > 10) { > + /* max total time: 1/2 * 100 * 2000 us = 1 second */ > + usleep(iterations * 2000); > + } > + for (offset = 0; offset < len; offset += PAGE_SIZE) { > + dummy = ((volatile unsigned char*)addr)[offset]; > + } > + } > return e; > } > > @@ -230,7 +257,7 @@ > pthread_key_create(&hcall_buf_pkey, _xc_clean_hcall_buf); > } > > -int hcall_buf_prep(void **addr, size_t len) > +int hcall_buf_prep(xc_interface *xch, void **addr, size_t len) > { > struct hcall_buf *hcall_buf; > > @@ -248,7 +275,7 @@ > if ( !hcall_buf->buf ) > { > hcall_buf->buf = xc_memalign(PAGE_SIZE, PAGE_SIZE); > - if ( !hcall_buf->buf || lock_pages(hcall_buf->buf, PAGE_SIZE) ) > + if ( !hcall_buf->buf || xc_lock_pages(xch, hcall_buf->buf, PAGE_SIZE) ) > { > free(hcall_buf->buf); > hcall_buf->buf = NULL; > @@ -265,7 +292,7 @@ > } > > out: > - return lock_pages(*addr, len); > + return xc_lock_pages(xch, *addr, len); > } > > void hcall_buf_release(void **addr, size_t len) > @@ -307,7 +334,7 @@ > DECLARE_HYPERCALL; > long ret = -EINVAL; > > - if ( hcall_buf_prep((void **)&op, nr_ops*sizeof(*op)) != 0 ) > + if ( hcall_buf_prep(xch, (void **)&op, nr_ops*sizeof(*op)) != 0 ) > { > PERROR("Could not lock memory for Xen hypercall"); > goto out1; > @@ -341,7 +368,7 @@ > hypercall.arg[2] = 0; > hypercall.arg[3] = mmu->subject; > > - if ( lock_pages(mmu->updates, sizeof(mmu->updates)) != 0 ) > + if ( xc_lock_pages(xch, mmu->updates, sizeof(mmu->updates)) != 0 ) > { > PERROR("flush_mmu_updates: mmu updates lock_pages failed"); > err = 1; > @@ -408,14 +435,14 @@ > case XENMEM_increase_reservation: > case XENMEM_decrease_reservation: > case XENMEM_populate_physmap: > - if ( lock_pages(reservation, sizeof(*reservation)) != 0 ) > + if ( xc_lock_pages(xch, reservation, sizeof(*reservation)) != 0 ) > { > PERROR("Could not lock"); > goto out1; > } > get_xen_guest_handle(extent_start, reservation->extent_start); > if ( (extent_start != NULL) && > - (lock_pages(extent_start, > + (xc_lock_pages(xch, extent_start, > reservation->nr_extents * sizeof(xen_pfn_t)) != 0) ) > { > PERROR("Could not lock"); > @@ -424,13 +451,13 @@ > } > break; > case XENMEM_machphys_mfn_list: > - if ( lock_pages(xmml, sizeof(*xmml)) != 0 ) > + if ( xc_lock_pages(xch, xmml, sizeof(*xmml)) != 0 ) > { > PERROR("Could not lock"); > goto out1; > } > get_xen_guest_handle(extent_start, xmml->extent_start); > - if ( lock_pages(extent_start, > + if ( xc_lock_pages(xch, extent_start, > xmml->max_extents * sizeof(xen_pfn_t)) != 0 ) > { > PERROR("Could not lock"); > @@ -439,7 +466,7 @@ > } > break; > case XENMEM_add_to_physmap: > - if ( lock_pages(arg, sizeof(struct xen_add_to_physmap)) ) > + if ( xc_lock_pages(xch, arg, sizeof(struct xen_add_to_physmap)) ) > { > PERROR("Could not lock"); > goto out1; > @@ -448,7 +475,7 @@ > case XENMEM_current_reservation: > case XENMEM_maximum_reservation: > case XENMEM_maximum_gpfn: > - if ( lock_pages(arg, sizeof(domid_t)) ) > + if ( xc_lock_pages(xch, arg, sizeof(domid_t)) ) > { > PERROR("Could not lock"); > goto out1; > @@ -456,7 +483,7 @@ > break; > case XENMEM_set_pod_target: > case XENMEM_get_pod_target: > - if ( lock_pages(arg, sizeof(struct xen_pod_target)) ) > + if ( xc_lock_pages(xch, arg, sizeof(struct xen_pod_target)) ) > { > PERROR("Could not lock"); > goto out1; > @@ -535,7 +562,7 @@ > memset(pfn_buf, 0, max_pfns * sizeof(*pfn_buf)); > #endif > > - if ( lock_pages(pfn_buf, max_pfns * sizeof(*pfn_buf)) != 0 ) > + if ( xc_lock_pages(xch, pfn_buf, max_pfns * sizeof(*pfn_buf)) != 0 ) > { > PERROR("xc_get_pfn_list: pfn_buf lock failed"); > return -1; > @@ -618,7 +645,7 @@ > break; > } > > - if ( (argsize != 0) && (lock_pages(arg, argsize) != 0) ) > + if ( (argsize != 0) && (xc_lock_pages(xch, arg, argsize) != 0) ) > { > PERROR("Could not lock memory for version hypercall"); > return -ENOMEM; > diff -r 267ecb2ee5bf tools/libxc/xc_private.h > --- a/tools/libxc/xc_private.h Tue Jun 01 14:25:59 2010 +0100 > +++ b/tools/libxc/xc_private.h Tue Jun 01 17:54:51 2010 +0100 > @@ -85,10 +85,10 @@ > > void *xc_memalign(size_t alignment, size_t size); > > -int lock_pages(void *addr, size_t len); > +int xc_lock_pages(xc_interface*, void *addr, size_t len); > void unlock_pages(void *addr, size_t len); > > -int hcall_buf_prep(void **addr, size_t len); > +int hcall_buf_prep(xc_interface*, void **addr, size_t len); > void hcall_buf_release(void **addr, size_t len); > > static inline void safe_munlock(const void *addr, size_t len) > @@ -117,7 +117,7 @@ > > DECLARE_HYPERCALL; > > - if ( hcall_buf_prep(&op, len) != 0 ) > + if ( hcall_buf_prep(xch, &op, len) != 0 ) > { > PERROR("Could not lock memory for Xen hypercall"); > goto out1; > @@ -145,7 +145,7 @@ > int ret = -1; > DECLARE_HYPERCALL; > > - if ( hcall_buf_prep((void **)&domctl, sizeof(*domctl)) != 0 ) > + if ( hcall_buf_prep(xch, (void **)&domctl, sizeof(*domctl)) != 0 ) > { > PERROR("Could not lock memory for Xen hypercall"); > goto out1; > @@ -174,7 +174,7 @@ > int ret = -1; > DECLARE_HYPERCALL; > > - if ( hcall_buf_prep((void **)&sysctl, sizeof(*sysctl)) != 0 ) > + if ( hcall_buf_prep(xch, (void **)&sysctl, sizeof(*sysctl)) != 0 ) > { > PERROR("Could not lock memory for Xen hypercall"); > goto out1; > diff -r 267ecb2ee5bf tools/libxc/xc_resume.c > --- a/tools/libxc/xc_resume.c Tue Jun 01 14:25:59 2010 +0100 > +++ b/tools/libxc/xc_resume.c Tue Jun 01 17:54:51 2010 +0100 > @@ -180,7 +180,7 @@ > goto out; > } > > - if ( lock_pages(&ctxt, sizeof(ctxt)) ) > + if ( xc_lock_pages(xch, &ctxt, sizeof(ctxt)) ) > { > ERROR("Unable to lock ctxt"); > goto out; > diff -r 267ecb2ee5bf tools/libxc/xc_tbuf.c > --- a/tools/libxc/xc_tbuf.c Tue Jun 01 14:25:59 2010 +0100 > +++ b/tools/libxc/xc_tbuf.c Tue Jun 01 17:54:51 2010 +0100 > @@ -120,7 +120,7 @@ > set_xen_guest_handle(sysctl.u.tbuf_op.cpu_mask.bitmap, bytemap); > sysctl.u.tbuf_op.cpu_mask.nr_cpus = sizeof(bytemap) * 8; > > - if ( lock_pages(&bytemap, sizeof(bytemap)) != 0 ) > + if ( xc_lock_pages(xch, &bytemap, sizeof(bytemap)) != 0 ) > { > PERROR("Could not lock memory for Xen hypercall"); > goto out; > diff -r 267ecb2ee5bf tools/libxc/xc_tmem.c > --- a/tools/libxc/xc_tmem.c Tue Jun 01 14:25:59 2010 +0100 > +++ b/tools/libxc/xc_tmem.c Tue Jun 01 17:54:51 2010 +0100 > @@ -14,7 +14,7 @@ > > hypercall.op = __HYPERVISOR_tmem_op; > hypercall.arg[0] = (unsigned long)op; > - if (lock_pages(op, sizeof(*op)) != 0) > + if (xc_lock_pages(xch, op, sizeof(*op)) != 0) > { > PERROR("Could not lock memory for Xen hypercall"); > return -EFAULT; > @@ -52,7 +52,7 @@ > op.u.ctrl.arg3 = arg3; > > if (subop == TMEMC_LIST) { > - if ((arg1 != 0) && (lock_pages(buf, arg1) != 0)) > + if ((arg1 != 0) && (xc_lock_pages(xch, buf, arg1) != 0)) > { > PERROR("Could not lock memory for Xen hypercall"); > return -ENOMEM; > ^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [xl] a fault happens on pv domain reboot 2010-06-02 9:10 ` Sergey Tovpeko @ 2010-06-02 9:27 ` Keir Fraser 2010-06-02 10:11 ` Ian Jackson 1 sibling, 0 replies; 9+ messages in thread From: Keir Fraser @ 2010-06-02 9:27 UTC (permalink / raw) To: Sergey Tovpeko, Ian Jackson; +Cc: xen-devel@lists.xensource.com, Stabellini Finding out which syscall involved in libxl_domain_info returns EAGAIN would be the obvious first step. It is probably one of the mlocks, since the hypercall itself should not be returning EAGAIN (if it is, then that's going to be a nice easy bug fix!). Root-causing why mlock is so often failing on your setup is going to be harder to debug, no doubt. -- Keir On 02/06/2010 10:10, "Sergey Tovpeko" <tsv.devel@gmail.com> wrote: > The mlock issue isn't my case. I mean the patch didn't solve the problem. > The mentioned mlock returned success with the first iteration. > > I thought below patch (trick) solves the issue, but it only decreases > the probability of the bug. > > diff -r 458589ad3793 tools/libxl/xl_cmdimpl.c > --- a/tools/libxl/xl_cmdimpl.c Tue Jun 01 07:06:50 2010 +0100 > +++ b/tools/libxl/xl_cmdimpl.c Wed Jun 02 16:57:59 2010 +0400 > @@ -1159,6 +1159,15 @@ > libxl_free_waiter(w2); > free(w1); > free(w2); > + while (1) { > + struct libxl_dominfo info_t; > + if (libxl_domain_info(&ctx, &info_t, > domid)) > + break; > + sleep(1); > + } > > LOG("Done. Rebooting now"); > goto start; > } > > >> Stefano Stabellini writes ("Re: [Xen-devel] [xl] a fault happens on pv domain >> reboot"): >> >>> That means that xc_domain_getinfolist failed with EAGAIN, the reason >>> seems to be that mlock can actually failed with EAGAIN but we don't do >>> anything about it in libxc. >>> If mlock returns EAGAIN we should make sure that the pages are actually >>> available in RAM and try again. >>> >> >> So, on that note, could you please try this patch ? >> >> Ian. >> >> diff -r 267ecb2ee5bf tools/libxc/xc_acm.c >> --- a/tools/libxc/xc_acm.c Tue Jun 01 14:25:59 2010 +0100 >> +++ b/tools/libxc/xc_acm.c Tue Jun 01 17:54:51 2010 +0100 >> @@ -83,7 +83,7 @@ >> >> hypercall.op = __HYPERVISOR_xsm_op; >> hypercall.arg[0] = (unsigned long)&acmctl; >> - if ( lock_pages(&acmctl, sizeof(acmctl)) != 0) >> + if ( xc_lock_pages(xch, &acmctl, sizeof(acmctl)) != 0) >> { >> PERROR("Could not lock memory for Xen hypercall"); >> return -EFAULT; >> diff -r 267ecb2ee5bf tools/libxc/xc_cpupool.c >> --- a/tools/libxc/xc_cpupool.c Tue Jun 01 14:25:59 2010 +0100 >> +++ b/tools/libxc/xc_cpupool.c Tue Jun 01 17:54:51 2010 +0100 >> @@ -71,7 +71,7 @@ >> set_xen_guest_handle(sysctl.u.cpupool_op.cpumap.bitmap, local); >> sysctl.u.cpupool_op.cpumap.nr_cpus = sizeof(info->cpumap) * 8; >> >> - if ( (err = lock_pages(local, sizeof(local))) != 0 ) >> + if ( (err = xc_lock_pages(xch, local, sizeof(local))) != 0 ) >> { >> PERROR("Could not lock memory for Xen hypercall"); >> break; >> @@ -147,7 +147,7 @@ >> set_xen_guest_handle(sysctl.u.cpupool_op.cpumap.bitmap, local); >> sysctl.u.cpupool_op.cpumap.nr_cpus = sizeof(*cpumap) * 8; >> >> - if ( (err = lock_pages(local, sizeof(local))) != 0 ) >> + if ( (err = xc_lock_pages(xch, local, sizeof(local))) != 0 ) >> { >> PERROR("Could not lock memory for Xen hypercall"); >> return err; >> diff -r 267ecb2ee5bf tools/libxc/xc_domain.c >> --- a/tools/libxc/xc_domain.c Tue Jun 01 14:25:59 2010 +0100 >> +++ b/tools/libxc/xc_domain.c Tue Jun 01 17:54:51 2010 +0100 >> @@ -80,7 +80,7 @@ >> arg.domain_id = domid; >> arg.reason = reason; >> >> - if ( lock_pages(&arg, sizeof(arg)) != 0 ) >> + if ( xc_lock_pages(xch, &arg, sizeof(arg)) != 0 ) >> { >> PERROR("Could not lock memory for Xen hypercall"); >> goto out1; >> @@ -119,7 +119,7 @@ >> >> domctl.u.vcpuaffinity.cpumap.nr_cpus = cpusize * 8; >> >> - if ( lock_pages(local, cpusize) != 0 ) >> + if ( xc_lock_pages(xch, local, cpusize) != 0 ) >> { >> PERROR("Could not lock memory for Xen hypercall"); >> goto out; >> @@ -158,7 +158,7 @@ >> set_xen_guest_handle(domctl.u.vcpuaffinity.cpumap.bitmap, local); >> domctl.u.vcpuaffinity.cpumap.nr_cpus = cpusize * 8; >> >> - if ( lock_pages(local, sizeof(local)) != 0 ) >> + if ( xc_lock_pages(xch, local, sizeof(local)) != 0 ) >> { >> PERROR("Could not lock memory for Xen hypercall"); >> goto out; >> @@ -243,7 +243,7 @@ >> int ret = 0; >> DECLARE_SYSCTL; >> >> - if ( lock_pages(info, max_domains*sizeof(xc_domaininfo_t)) != 0 ) >> + if ( xc_lock_pages(xch, info, max_domains*sizeof(xc_domaininfo_t)) != 0 >> ) >> return -1; >> >> sysctl.cmd = XEN_SYSCTL_getdomaininfolist; >> @@ -276,7 +276,7 @@ >> set_xen_guest_handle(domctl.u.hvmcontext.buffer, ctxt_buf); >> >> if ( ctxt_buf ) >> - if ( (ret = lock_pages(ctxt_buf, size)) != 0 ) >> + if ( (ret = xc_lock_pages(xch, ctxt_buf, size)) != 0 ) >> return ret; >> >> ret = do_domctl(xch, &domctl); >> @@ -308,7 +308,7 @@ >> domctl.u.hvmcontext_partial.instance = instance; >> set_xen_guest_handle(domctl.u.hvmcontext_partial.buffer, ctxt_buf); >> >> - if ( (ret = lock_pages(ctxt_buf, size)) != 0 ) >> + if ( (ret = xc_lock_pages(xch, ctxt_buf, size)) != 0 ) >> return ret; >> >> ret = do_domctl(xch, &domctl); >> @@ -333,7 +333,7 @@ >> domctl.u.hvmcontext.size = size; >> set_xen_guest_handle(domctl.u.hvmcontext.buffer, ctxt_buf); >> >> - if ( (ret = lock_pages(ctxt_buf, size)) != 0 ) >> + if ( (ret = xc_lock_pages(xch, ctxt_buf, size)) != 0 ) >> return ret; >> >> ret = do_domctl(xch, &domctl); >> @@ -358,7 +358,7 @@ >> set_xen_guest_handle(domctl.u.vcpucontext.ctxt, &ctxt->c); >> >> >> - if ( (rc = lock_pages(ctxt, sz)) != 0 ) >> + if ( (rc = xc_lock_pages(xch, ctxt, sz)) != 0 ) >> return rc; >> rc = do_domctl(xch, &domctl); >> unlock_pages(ctxt, sz); >> @@ -446,7 +446,8 @@ >> >> set_xen_guest_handle(fmap.map.buffer, &e820); >> >> - if ( lock_pages(&fmap, sizeof(fmap)) || lock_pages(&e820, sizeof(e820)) >> ) >> + if ( xc_lock_pages(xch, &fmap, sizeof(fmap)) || >> + xc_lock_pages(xch, &e820, sizeof(e820)) ) >> { >> PERROR("Could not lock memory for Xen hypercall"); >> rc = -1; >> @@ -522,7 +523,7 @@ >> domctl.cmd = XEN_DOMCTL_gettscinfo; >> domctl.domain = (domid_t)domid; >> set_xen_guest_handle(domctl.u.tsc_info.out_info, &info); >> - if ( (rc = lock_pages(&info, sizeof(info))) != 0 ) >> + if ( (rc = xc_lock_pages(xch, &info, sizeof(info))) != 0 ) >> return rc; >> rc = do_domctl(xch, &domctl); >> if ( rc == 0 ) >> @@ -807,7 +808,7 @@ >> domctl.u.vcpucontext.vcpu = vcpu; >> set_xen_guest_handle(domctl.u.vcpucontext.ctxt, &ctxt->c); >> >> - if ( (rc = lock_pages(ctxt, sz)) != 0 ) >> + if ( (rc = xc_lock_pages(xch, ctxt, sz)) != 0 ) >> return rc; >> rc = do_domctl(xch, &domctl); >> >> @@ -875,7 +876,7 @@ >> arg.domid = dom; >> arg.index = param; >> arg.value = value; >> - if ( lock_pages(&arg, sizeof(arg)) != 0 ) >> + if ( xc_lock_pages(handle, &arg, sizeof(arg)) != 0 ) >> return -1; >> rc = do_xen_hypercall(handle, &hypercall); >> unlock_pages(&arg, sizeof(arg)); >> @@ -893,7 +894,7 @@ >> hypercall.arg[1] = (unsigned long)&arg; >> arg.domid = dom; >> arg.index = param; >> - if ( lock_pages(&arg, sizeof(arg)) != 0 ) >> + if ( xc_lock_pages(handle, &arg, sizeof(arg)) != 0 ) >> return -1; >> rc = do_xen_hypercall(handle, &hypercall); >> unlock_pages(&arg, sizeof(arg)); >> @@ -946,7 +947,7 @@ >> >> set_xen_guest_handle(domctl.u.get_device_group.sdev_array, sdev_array); >> >> - if ( lock_pages(sdev_array, max_sdevs * sizeof(*sdev_array)) != 0 ) >> + if ( xc_lock_pages(xch, sdev_array, max_sdevs * sizeof(*sdev_array)) != >> 0 ) >> { >> PERROR("Could not lock memory for xc_get_device_group"); >> return -ENOMEM; >> diff -r 267ecb2ee5bf tools/libxc/xc_domain_restore.c >> --- a/tools/libxc/xc_domain_restore.c Tue Jun 01 14:25:59 2010 +0100 >> +++ b/tools/libxc/xc_domain_restore.c Tue Jun 01 17:54:51 2010 +0100 >> @@ -1451,7 +1451,7 @@ >> memset(region_mfn, 0, >> ROUNDUP(MAX_BATCH_SIZE * sizeof(xen_pfn_t), PAGE_SHIFT)); >> >> - if ( lock_pages(region_mfn, sizeof(xen_pfn_t) * MAX_BATCH_SIZE) ) >> + if ( xc_lock_pages(xch, region_mfn, sizeof(xen_pfn_t) * MAX_BATCH_SIZE) >> ) >> { >> PERROR("Could not lock region_mfn"); >> goto out; >> @@ -1801,7 +1801,7 @@ >> } >> } >> >> - if ( lock_pages(&ctxt, sizeof(ctxt)) ) >> + if ( xc_lock_pages(xch, &ctxt, sizeof(ctxt)) ) >> { >> PERROR("Unable to lock ctxt"); >> return 1; >> diff -r 267ecb2ee5bf tools/libxc/xc_domain_save.c >> --- a/tools/libxc/xc_domain_save.c Tue Jun 01 14:25:59 2010 +0100 >> +++ b/tools/libxc/xc_domain_save.c Tue Jun 01 17:54:51 2010 +0100 >> @@ -1032,14 +1032,14 @@ >> >> memset(to_send, 0xff, BITMAP_SIZE); >> >> - if ( lock_pages(to_send, BITMAP_SIZE) ) >> + if ( xc_lock_pages(xch, to_send, BITMAP_SIZE) ) >> { >> PERROR("Unable to lock to_send"); >> return 1; >> } >> >> /* (to fix is local only) */ >> - if ( lock_pages(to_skip, BITMAP_SIZE) ) >> + if ( xc_lock_pages(xch, to_skip, BITMAP_SIZE) ) >> { >> PERROR("Unable to lock to_skip"); >> return 1; >> @@ -1077,7 +1077,7 @@ >> memset(pfn_type, 0, >> ROUNDUP(MAX_BATCH_SIZE * sizeof(*pfn_type), PAGE_SHIFT)); >> >> - if ( lock_pages(pfn_type, MAX_BATCH_SIZE * sizeof(*pfn_type)) ) >> + if ( xc_lock_pages(xch, pfn_type, MAX_BATCH_SIZE * sizeof(*pfn_type)) ) >> { >> PERROR("Unable to lock pfn_type array"); >> goto out; >> diff -r 267ecb2ee5bf tools/libxc/xc_evtchn.c >> --- a/tools/libxc/xc_evtchn.c Tue Jun 01 14:25:59 2010 +0100 >> +++ b/tools/libxc/xc_evtchn.c Tue Jun 01 17:54:51 2010 +0100 >> @@ -19,7 +19,7 @@ >> hypercall.arg[0] = cmd; >> hypercall.arg[1] = (unsigned long)arg; >> >> - if ( lock_pages(arg, arg_size) != 0 ) >> + if ( xc_lock_pages(xch, arg, arg_size) != 0 ) >> { >> PERROR("do_evtchn_op: arg lock failed"); >> goto out; >> diff -r 267ecb2ee5bf tools/libxc/xc_linux.c >> --- a/tools/libxc/xc_linux.c Tue Jun 01 14:25:59 2010 +0100 >> +++ b/tools/libxc/xc_linux.c Tue Jun 01 17:54:51 2010 +0100 >> @@ -724,7 +724,7 @@ >> hypercall.arg[1] = (unsigned long)op; >> hypercall.arg[2] = count; >> >> - if ( lock_pages(op, count* op_size) != 0 ) >> + if ( xc_lock_pages(xch, op, count* op_size) != 0 ) >> { >> PERROR("Could not lock memory for Xen hypercall"); >> goto out1; >> @@ -776,7 +776,7 @@ >> *gnt_num = query.nr_frames * (PAGE_SIZE / sizeof(grant_entry_v1_t) ); >> >> frame_list = malloc(query.nr_frames * sizeof(unsigned long)); >> - if ( !frame_list || lock_pages(frame_list, >> + if ( !frame_list || xc_lock_pages(xch, frame_list, >> query.nr_frames * sizeof(unsigned long)) >> ) >> { >> ERROR("Alloc/lock frame_list in xc_gnttab_map_table\n"); >> diff -r 267ecb2ee5bf tools/libxc/xc_misc.c >> --- a/tools/libxc/xc_misc.c Tue Jun 01 14:25:59 2010 +0100 >> +++ b/tools/libxc/xc_misc.c Tue Jun 01 17:54:51 2010 +0100 >> @@ -28,7 +28,7 @@ >> sysctl.u.readconsole.incremental = incremental; >> } >> >> - if ( (ret = lock_pages(buffer, nr_chars)) != 0 ) >> + if ( (ret = xc_lock_pages(xch, buffer, nr_chars)) != 0 ) >> return ret; >> >> if ( (ret = do_sysctl(xch, &sysctl)) == 0 ) >> @@ -52,7 +52,7 @@ >> set_xen_guest_handle(sysctl.u.debug_keys.keys, keys); >> sysctl.u.debug_keys.nr_keys = len; >> >> - if ( (ret = lock_pages(keys, len)) != 0 ) >> + if ( (ret = xc_lock_pages(xch, keys, len)) != 0 ) >> return ret; >> >> ret = do_sysctl(xch, &sysctl); >> @@ -140,7 +140,7 @@ >> DECLARE_HYPERCALL; >> >> mc->interface_version = XEN_MCA_INTERFACE_VERSION; >> - if ( lock_pages(mc, sizeof(mc)) ) >> + if ( xc_lock_pages(xch, mc, sizeof(mc)) ) >> { >> PERROR("Could not lock xen_mc memory"); >> return -EINVAL; >> @@ -213,7 +213,7 @@ >> sysctl.u.getcpuinfo.max_cpus = max_cpus; >> set_xen_guest_handle(sysctl.u.getcpuinfo.info, info); >> >> - if ( (rc = lock_pages(info, max_cpus*sizeof(*info))) != 0 ) >> + if ( (rc = xc_lock_pages(xch, info, max_cpus*sizeof(*info))) != 0 ) >> return rc; >> >> rc = do_sysctl(xch, &sysctl); >> @@ -236,7 +236,7 @@ >> struct xen_hvm_set_pci_intx_level _arg, *arg = &_arg; >> int rc; >> >> - if ( (rc = hcall_buf_prep((void **)&arg, sizeof(*arg))) != 0 ) >> + if ( (rc = hcall_buf_prep(xch, (void **)&arg, sizeof(*arg))) != 0 ) >> { >> PERROR("Could not lock memory"); >> return rc; >> @@ -269,7 +269,7 @@ >> struct xen_hvm_set_isa_irq_level _arg, *arg = &_arg; >> int rc; >> >> - if ( (rc = hcall_buf_prep((void **)&arg, sizeof(*arg))) != 0 ) >> + if ( (rc = hcall_buf_prep(xch, (void **)&arg, sizeof(*arg))) != 0 ) >> { >> PERROR("Could not lock memory"); >> return rc; >> @@ -305,7 +305,7 @@ >> arg.link = link; >> arg.isa_irq = isa_irq; >> >> - if ( (rc = lock_pages(&arg, sizeof(arg))) != 0 ) >> + if ( (rc = xc_lock_pages(xch, &arg, sizeof(arg))) != 0 ) >> { >> PERROR("Could not lock memory"); >> return rc; >> @@ -336,7 +336,7 @@ >> arg.nr = nr; >> set_xen_guest_handle(arg.dirty_bitmap, (uint8_t *)dirty_bitmap); >> >> - if ( (rc = lock_pages(&arg, sizeof(arg))) != 0 ) >> + if ( (rc = xc_lock_pages(xch, &arg, sizeof(arg))) != 0 ) >> { >> PERROR("Could not lock memory"); >> return rc; >> @@ -364,7 +364,7 @@ >> arg.first_pfn = first_pfn; >> arg.nr = nr; >> >> - if ( (rc = lock_pages(&arg, sizeof(arg))) != 0 ) >> + if ( (rc = xc_lock_pages(xch, &arg, sizeof(arg))) != 0 ) >> { >> PERROR("Could not lock memory"); >> return rc; >> @@ -393,7 +393,7 @@ >> arg.first_pfn = first_pfn; >> arg.nr = nr; >> >> - if ( (rc = lock_pages(&arg, sizeof(arg))) != 0 ) >> + if ( (rc = xc_lock_pages(xch, &arg, sizeof(arg))) != 0 ) >> { >> PERROR("Could not lock memory"); >> return rc; >> diff -r 267ecb2ee5bf tools/libxc/xc_offline_page.c >> --- a/tools/libxc/xc_offline_page.c Tue Jun 01 14:25:59 2010 +0100 >> +++ b/tools/libxc/xc_offline_page.c Tue Jun 01 17:54:51 2010 +0100 >> @@ -57,7 +57,7 @@ >> if ( !status || (end < start) ) >> return -EINVAL; >> >> - if (lock_pages(status, sizeof(uint32_t)*(end - start + 1))) >> + if (xc_lock_pages(xch, status, sizeof(uint32_t)*(end - start + 1))) >> { >> ERROR("Could not lock memory for xc_mark_page_online\n"); >> return -EINVAL; >> @@ -84,7 +84,7 @@ >> if ( !status || (end < start) ) >> return -EINVAL; >> >> - if (lock_pages(status, sizeof(uint32_t)*(end - start + 1))) >> + if (xc_lock_pages(xch, status, sizeof(uint32_t)*(end - start + 1))) >> { >> ERROR("Could not lock memory for xc_mark_page_offline"); >> return -EINVAL; >> @@ -111,7 +111,7 @@ >> if ( !status || (end < start) ) >> return -EINVAL; >> >> - if (lock_pages(status, sizeof(uint32_t)*(end - start + 1))) >> + if (xc_lock_pages(xch, status, sizeof(uint32_t)*(end - start + 1))) >> { >> ERROR("Could not lock memory for xc_query_page_offline_status\n"); >> return -EINVAL; >> @@ -277,7 +277,7 @@ >> minfo->pfn_type[i] = pfn_to_mfn(i, minfo->p2m_table, >> minfo->guest_width); >> >> - if ( lock_pages(minfo->pfn_type, minfo->p2m_size * >> sizeof(*minfo->pfn_type)) ) >> + if ( xc_lock_pages(xch, minfo->pfn_type, minfo->p2m_size * >> sizeof(*minfo->pfn_type)) ) >> { >> ERROR("Unable to lock pfn_type array"); >> goto failed; >> diff -r 267ecb2ee5bf tools/libxc/xc_pm.c >> --- a/tools/libxc/xc_pm.c Tue Jun 01 14:25:59 2010 +0100 >> +++ b/tools/libxc/xc_pm.c Tue Jun 01 17:54:51 2010 +0100 >> @@ -57,11 +57,11 @@ >> if ( (ret = xc_pm_get_max_px(xch, cpuid, &max_px)) != 0) >> return ret; >> >> - if ( (ret = lock_pages(pxpt->trans_pt, >> + if ( (ret = xc_lock_pages(xch, pxpt->trans_pt, >> max_px * max_px * sizeof(uint64_t))) != 0 ) >> return ret; >> >> - if ( (ret = lock_pages(pxpt->pt, >> + if ( (ret = xc_lock_pages(xch, pxpt->pt, >> max_px * sizeof(struct xc_px_val))) != 0 ) >> { >> unlock_pages(pxpt->trans_pt, max_px * max_px * sizeof(uint64_t)); >> @@ -132,11 +132,11 @@ >> if ( (ret = xc_pm_get_max_cx(xch, cpuid, &max_cx)) ) >> goto unlock_0; >> >> - if ( (ret = lock_pages(cxpt, sizeof(struct xc_cx_stat))) ) >> + if ( (ret = xc_lock_pages(xch, cxpt, sizeof(struct xc_cx_stat))) ) >> goto unlock_0; >> - if ( (ret = lock_pages(cxpt->triggers, max_cx * sizeof(uint64_t))) ) >> + if ( (ret = xc_lock_pages(xch, cxpt->triggers, max_cx * >> sizeof(uint64_t))) ) >> goto unlock_1; >> - if ( (ret = lock_pages(cxpt->residencies, max_cx * sizeof(uint64_t))) ) >> + if ( (ret = xc_lock_pages(xch, cxpt->residencies, max_cx * >> sizeof(uint64_t))) ) >> goto unlock_2; >> >> sysctl.cmd = XEN_SYSCTL_get_pmstat; >> @@ -199,13 +199,13 @@ >> (!user_para->scaling_available_governors) ) >> return -EINVAL; >> >> - if ( (ret = lock_pages(user_para->affected_cpus, >> + if ( (ret = xc_lock_pages(xch, user_para->affected_cpus, >> user_para->cpu_num * sizeof(uint32_t))) ) >> goto unlock_1; >> - if ( (ret = lock_pages(user_para->scaling_available_frequencies, >> + if ( (ret = xc_lock_pages(xch, >> user_para->scaling_available_frequencies, >> user_para->freq_num * sizeof(uint32_t))) ) >> goto unlock_2; >> - if ( (ret = lock_pages(user_para->scaling_available_governors, >> + if ( (ret = xc_lock_pages(xch, >> user_para->scaling_available_governors, >> user_para->gov_num * CPUFREQ_NAME_LEN * sizeof(char))) ) >> goto unlock_3; >> >> diff -r 267ecb2ee5bf tools/libxc/xc_private.c >> --- a/tools/libxc/xc_private.c Tue Jun 01 14:25:59 2010 +0100 >> +++ b/tools/libxc/xc_private.c Tue Jun 01 17:54:51 2010 +0100 >> @@ -174,7 +174,7 @@ >> >> #ifdef __sun__ >> >> -int lock_pages(void *addr, size_t len) { return 0; } >> +int xc_lock_pages(xc_interface *xch, void *addr, size_t len) { return 0; } >> void unlock_pages(void *addr, size_t len) { } >> >> int hcall_buf_prep(void **addr, size_t len) { return 0; } >> @@ -182,13 +182,40 @@ >> >> #else /* !__sun__ */ >> >> -int lock_pages(void *addr, size_t len) >> +int xc_lock_pages(xc_interface *xch, void *addr, size_t len) >> { >> int e; >> void *laddr = (void *)((unsigned long)addr & PAGE_MASK); >> size_t llen = (len + ((unsigned long)addr - (unsigned long)laddr) + >> PAGE_SIZE - 1) & PAGE_MASK; >> - e = mlock(laddr, llen); >> + size_t offset; >> + int iterations = 0; >> + int dummy; >> + >> + for (;;) { >> + e = mlock(laddr, llen); >> + if (!e) { >> + if (iterations > 5) >> + DBGPRINTF("mlock (libxc_lock_pages) (len=%zu)" >> + " took %d iterations", len, iterations); >> + return 0; >> + } >> + if (errno != EAGAIN) { >> + PERROR("mlock (libxc_lock_pages) failed (len=%zu)", len); >> + return e; >> + } >> + if (++iterations > 100) { >> + ERROR("mlock (libxc_lock_pages) too much EAGAIN (len=%zu)", >> len); >> + return -1; >> + } >> + if (iterations > 10) { >> + /* max total time: 1/2 * 100 * 2000 us = 1 second */ >> + usleep(iterations * 2000); >> + } >> + for (offset = 0; offset < len; offset += PAGE_SIZE) { >> + dummy = ((volatile unsigned char*)addr)[offset]; >> + } >> + } >> return e; >> } >> >> @@ -230,7 +257,7 @@ >> pthread_key_create(&hcall_buf_pkey, _xc_clean_hcall_buf); >> } >> >> -int hcall_buf_prep(void **addr, size_t len) >> +int hcall_buf_prep(xc_interface *xch, void **addr, size_t len) >> { >> struct hcall_buf *hcall_buf; >> >> @@ -248,7 +275,7 @@ >> if ( !hcall_buf->buf ) >> { >> hcall_buf->buf = xc_memalign(PAGE_SIZE, PAGE_SIZE); >> - if ( !hcall_buf->buf || lock_pages(hcall_buf->buf, PAGE_SIZE) ) >> + if ( !hcall_buf->buf || xc_lock_pages(xch, hcall_buf->buf, >> PAGE_SIZE) ) >> { >> free(hcall_buf->buf); >> hcall_buf->buf = NULL; >> @@ -265,7 +292,7 @@ >> } >> >> out: >> - return lock_pages(*addr, len); >> + return xc_lock_pages(xch, *addr, len); >> } >> >> void hcall_buf_release(void **addr, size_t len) >> @@ -307,7 +334,7 @@ >> DECLARE_HYPERCALL; >> long ret = -EINVAL; >> >> - if ( hcall_buf_prep((void **)&op, nr_ops*sizeof(*op)) != 0 ) >> + if ( hcall_buf_prep(xch, (void **)&op, nr_ops*sizeof(*op)) != 0 ) >> { >> PERROR("Could not lock memory for Xen hypercall"); >> goto out1; >> @@ -341,7 +368,7 @@ >> hypercall.arg[2] = 0; >> hypercall.arg[3] = mmu->subject; >> >> - if ( lock_pages(mmu->updates, sizeof(mmu->updates)) != 0 ) >> + if ( xc_lock_pages(xch, mmu->updates, sizeof(mmu->updates)) != 0 ) >> { >> PERROR("flush_mmu_updates: mmu updates lock_pages failed"); >> err = 1; >> @@ -408,14 +435,14 @@ >> case XENMEM_increase_reservation: >> case XENMEM_decrease_reservation: >> case XENMEM_populate_physmap: >> - if ( lock_pages(reservation, sizeof(*reservation)) != 0 ) >> + if ( xc_lock_pages(xch, reservation, sizeof(*reservation)) != 0 ) >> { >> PERROR("Could not lock"); >> goto out1; >> } >> get_xen_guest_handle(extent_start, reservation->extent_start); >> if ( (extent_start != NULL) && >> - (lock_pages(extent_start, >> + (xc_lock_pages(xch, extent_start, >> reservation->nr_extents * sizeof(xen_pfn_t)) != 0) ) >> { >> PERROR("Could not lock"); >> @@ -424,13 +451,13 @@ >> } >> break; >> case XENMEM_machphys_mfn_list: >> - if ( lock_pages(xmml, sizeof(*xmml)) != 0 ) >> + if ( xc_lock_pages(xch, xmml, sizeof(*xmml)) != 0 ) >> { >> PERROR("Could not lock"); >> goto out1; >> } >> get_xen_guest_handle(extent_start, xmml->extent_start); >> - if ( lock_pages(extent_start, >> + if ( xc_lock_pages(xch, extent_start, >> xmml->max_extents * sizeof(xen_pfn_t)) != 0 ) >> { >> PERROR("Could not lock"); >> @@ -439,7 +466,7 @@ >> } >> break; >> case XENMEM_add_to_physmap: >> - if ( lock_pages(arg, sizeof(struct xen_add_to_physmap)) ) >> + if ( xc_lock_pages(xch, arg, sizeof(struct xen_add_to_physmap)) ) >> { >> PERROR("Could not lock"); >> goto out1; >> @@ -448,7 +475,7 @@ >> case XENMEM_current_reservation: >> case XENMEM_maximum_reservation: >> case XENMEM_maximum_gpfn: >> - if ( lock_pages(arg, sizeof(domid_t)) ) >> + if ( xc_lock_pages(xch, arg, sizeof(domid_t)) ) >> { >> PERROR("Could not lock"); >> goto out1; >> @@ -456,7 +483,7 @@ >> break; >> case XENMEM_set_pod_target: >> case XENMEM_get_pod_target: >> - if ( lock_pages(arg, sizeof(struct xen_pod_target)) ) >> + if ( xc_lock_pages(xch, arg, sizeof(struct xen_pod_target)) ) >> { >> PERROR("Could not lock"); >> goto out1; >> @@ -535,7 +562,7 @@ >> memset(pfn_buf, 0, max_pfns * sizeof(*pfn_buf)); >> #endif >> >> - if ( lock_pages(pfn_buf, max_pfns * sizeof(*pfn_buf)) != 0 ) >> + if ( xc_lock_pages(xch, pfn_buf, max_pfns * sizeof(*pfn_buf)) != 0 ) >> { >> PERROR("xc_get_pfn_list: pfn_buf lock failed"); >> return -1; >> @@ -618,7 +645,7 @@ >> break; >> } >> >> - if ( (argsize != 0) && (lock_pages(arg, argsize) != 0) ) >> + if ( (argsize != 0) && (xc_lock_pages(xch, arg, argsize) != 0) ) >> { >> PERROR("Could not lock memory for version hypercall"); >> return -ENOMEM; >> diff -r 267ecb2ee5bf tools/libxc/xc_private.h >> --- a/tools/libxc/xc_private.h Tue Jun 01 14:25:59 2010 +0100 >> +++ b/tools/libxc/xc_private.h Tue Jun 01 17:54:51 2010 +0100 >> @@ -85,10 +85,10 @@ >> >> void *xc_memalign(size_t alignment, size_t size); >> >> -int lock_pages(void *addr, size_t len); >> +int xc_lock_pages(xc_interface*, void *addr, size_t len); >> void unlock_pages(void *addr, size_t len); >> >> -int hcall_buf_prep(void **addr, size_t len); >> +int hcall_buf_prep(xc_interface*, void **addr, size_t len); >> void hcall_buf_release(void **addr, size_t len); >> >> static inline void safe_munlock(const void *addr, size_t len) >> @@ -117,7 +117,7 @@ >> >> DECLARE_HYPERCALL; >> >> - if ( hcall_buf_prep(&op, len) != 0 ) >> + if ( hcall_buf_prep(xch, &op, len) != 0 ) >> { >> PERROR("Could not lock memory for Xen hypercall"); >> goto out1; >> @@ -145,7 +145,7 @@ >> int ret = -1; >> DECLARE_HYPERCALL; >> >> - if ( hcall_buf_prep((void **)&domctl, sizeof(*domctl)) != 0 ) >> + if ( hcall_buf_prep(xch, (void **)&domctl, sizeof(*domctl)) != 0 ) >> { >> PERROR("Could not lock memory for Xen hypercall"); >> goto out1; >> @@ -174,7 +174,7 @@ >> int ret = -1; >> DECLARE_HYPERCALL; >> >> - if ( hcall_buf_prep((void **)&sysctl, sizeof(*sysctl)) != 0 ) >> + if ( hcall_buf_prep(xch, (void **)&sysctl, sizeof(*sysctl)) != 0 ) >> { >> PERROR("Could not lock memory for Xen hypercall"); >> goto out1; >> diff -r 267ecb2ee5bf tools/libxc/xc_resume.c >> --- a/tools/libxc/xc_resume.c Tue Jun 01 14:25:59 2010 +0100 >> +++ b/tools/libxc/xc_resume.c Tue Jun 01 17:54:51 2010 +0100 >> @@ -180,7 +180,7 @@ >> goto out; >> } >> >> - if ( lock_pages(&ctxt, sizeof(ctxt)) ) >> + if ( xc_lock_pages(xch, &ctxt, sizeof(ctxt)) ) >> { >> ERROR("Unable to lock ctxt"); >> goto out; >> diff -r 267ecb2ee5bf tools/libxc/xc_tbuf.c >> --- a/tools/libxc/xc_tbuf.c Tue Jun 01 14:25:59 2010 +0100 >> +++ b/tools/libxc/xc_tbuf.c Tue Jun 01 17:54:51 2010 +0100 >> @@ -120,7 +120,7 @@ >> set_xen_guest_handle(sysctl.u.tbuf_op.cpu_mask.bitmap, bytemap); >> sysctl.u.tbuf_op.cpu_mask.nr_cpus = sizeof(bytemap) * 8; >> >> - if ( lock_pages(&bytemap, sizeof(bytemap)) != 0 ) >> + if ( xc_lock_pages(xch, &bytemap, sizeof(bytemap)) != 0 ) >> { >> PERROR("Could not lock memory for Xen hypercall"); >> goto out; >> diff -r 267ecb2ee5bf tools/libxc/xc_tmem.c >> --- a/tools/libxc/xc_tmem.c Tue Jun 01 14:25:59 2010 +0100 >> +++ b/tools/libxc/xc_tmem.c Tue Jun 01 17:54:51 2010 +0100 >> @@ -14,7 +14,7 @@ >> >> hypercall.op = __HYPERVISOR_tmem_op; >> hypercall.arg[0] = (unsigned long)op; >> - if (lock_pages(op, sizeof(*op)) != 0) >> + if (xc_lock_pages(xch, op, sizeof(*op)) != 0) >> { >> PERROR("Could not lock memory for Xen hypercall"); >> return -EFAULT; >> @@ -52,7 +52,7 @@ >> op.u.ctrl.arg3 = arg3; >> >> if (subop == TMEMC_LIST) { >> - if ((arg1 != 0) && (lock_pages(buf, arg1) != 0)) >> + if ((arg1 != 0) && (xc_lock_pages(xch, buf, arg1) != 0)) >> { >> PERROR("Could not lock memory for Xen hypercall"); >> return -ENOMEM; >> > > > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.xensource.com > http://lists.xensource.com/xen-devel ^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [xl] a fault happens on pv domain reboot 2010-06-02 9:10 ` Sergey Tovpeko 2010-06-02 9:27 ` Keir Fraser @ 2010-06-02 10:11 ` Ian Jackson 2010-06-02 10:52 ` Sergey Tovpeko 1 sibling, 1 reply; 9+ messages in thread From: Ian Jackson @ 2010-06-02 10:11 UTC (permalink / raw) To: Sergey Tovpeko; +Cc: xen-devel@lists.xensource.com, Stefano Stabellini Sergey Tovpeko writes ("Re: [Xen-devel] [xl] a fault happens on pv domain reboot"): > The mlock issue isn't my case. I mean the patch didn't solve the problem. > The mentioned mlock returned success with the first iteration. So, just to be clear, with that 700-line patch it still failed with an identical error message ? Ian. ^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [xl] a fault happens on pv domain reboot 2010-06-02 10:11 ` Ian Jackson @ 2010-06-02 10:52 ` Sergey Tovpeko 2010-06-02 16:05 ` Ian Jackson 0 siblings, 1 reply; 9+ messages in thread From: Sergey Tovpeko @ 2010-06-02 10:52 UTC (permalink / raw) To: Ian Jackson; +Cc: xen-devel@lists.xensource.com, Stefano Stabellini Yes, the error message is the same. And i haven't noticed a message from the new xl_lock_page function. Waiting for domain debian-pv (domid 29) to die [pid 19628] Domain 29 is dead Domain 29 needs to be clean: destroying the domain Done. Rebooting now libxl: error: libxl_dom.c:389:userdata_path unable to find domain info for domain 30: Resource temporarily unavailable cannot save config file: Resource temporarily unavailable libxl: error: libxl.c:747:libxl_domain_destroy xc_domain_pause failed for 30 > Sergey Tovpeko writes ("Re: [Xen-devel] [xl] a fault happens on pv domain reboot"): > >> The mlock issue isn't my case. I mean the patch didn't solve the problem. >> The mentioned mlock returned success with the first iteration. >> > > So, just to be clear, with that 700-line patch it still failed with an > identical error message ? > > Ian. > ^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [xl] a fault happens on pv domain reboot 2010-06-02 10:52 ` Sergey Tovpeko @ 2010-06-02 16:05 ` Ian Jackson 0 siblings, 0 replies; 9+ messages in thread From: Ian Jackson @ 2010-06-02 16:05 UTC (permalink / raw) To: Sergey Tovpeko; +Cc: xen-devel@lists.xensource.com, Stabellini Sergey Tovpeko writes ("Re: [Xen-devel] [xl] a fault happens on pv domain reboot"): > Yes, the error message is the same. > And i haven't noticed a message from the new xl_lock_page function. Right, thanks. Do you know how to use strace ? > Waiting for domain debian-pv (domid 29) to die [pid 19628] If you run this again and at that point say strace -vvs500 -ostrace.out -p19628 substituting the right process for 19628, and then initiate the reboot, you should be able to see in strace.out what the failing system call is. Ian. ^ permalink raw reply [flat|nested] 9+ messages in thread
end of thread, other threads:[~2010-06-02 16:05 UTC | newest] Thread overview: 9+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2010-06-01 14:42 [xl] a fault happens on pv domain reboot Sergey Tovpeko 2010-06-01 14:53 ` Gihan Munasinghe 2010-06-01 16:05 ` Stefano Stabellini 2010-06-01 16:58 ` Ian Jackson 2010-06-02 9:10 ` Sergey Tovpeko 2010-06-02 9:27 ` Keir Fraser 2010-06-02 10:11 ` Ian Jackson 2010-06-02 10:52 ` Sergey Tovpeko 2010-06-02 16:05 ` Ian Jackson
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for NNTP newsgroup(s).