SeaBIOS July 2012

seabios@seabios.org

37 participants
119 discussions

[RFC PATCH v2 18/21] acpi_piix4: Update dimm bitmap state on hot-remove fail
by Vasilis Liaskovitis 11 Jul '12

11 Jul '12

This allows failed hot operations to be retried at anytime. This only works for guests that use _OST notification. Other guests cannot retry failed hot operations on same devices until after reboot. Signed-off-by: Vasilis Liaskovitis <vasilis.liaskovitis(a)profitbricks.com> --- hw/acpi_piix4.c | 20 +++++++++++++++++++- hw/dimm.c | 16 +++++++++++++++- hw/dimm.h | 2 +- 3 files changed, 35 insertions(+), 3 deletions(-) diff --git a/hw/acpi_piix4.c b/hw/acpi_piix4.c index ebc5de7..db631cc 100644 --- a/hw/acpi_piix4.c +++ b/hw/acpi_piix4.c @@ -599,6 +599,7 @@ static uint32_t pcirmv_read(void *opaque, uint32_t addr) static int piix4_device_hotplug(DeviceState *qdev, PCIDevice *dev, PCIHotplugState state); static int piix4_dimm_hotplug(DeviceState *qdev, SysBusDevice *dev, int add); +static int piix4_dimm_revert(DeviceState *qdev, SysBusDevice *dev, int add); static void piix4_acpi_system_hot_add_init(PCIBus *bus, PIIX4PMState *s) { @@ -627,7 +628,7 @@ static void piix4_acpi_system_hot_add_init(PCIBus *bus, PIIX4PMState *s) } pci_bus_hotplug(bus, piix4_device_hotplug, &s->dev.qdev); - dimm_register_hotplug(piix4_dimm_hotplug, &s->dev.qdev); + dimm_register_hotplug(piix4_dimm_hotplug, piix4_dimm_revert, &s->dev.qdev); } static void enable_device(PIIX4PMState *s, int slot) @@ -696,6 +697,23 @@ void piix4_dimm_state_sync(PIIX4PMState *s) } } +static int piix4_dimm_revert(DeviceState *qdev, SysBusDevice *dev, int add) +{ + PCIDevice *pci_dev = DO_UPCAST(PCIDevice, qdev, qdev); + PIIX4PMState *s = DO_UPCAST(PIIX4PMState, dev, pci_dev); + struct gpe_regs *g = &s->gperegs; + DimmState *slot = DIMM(dev); + int idx = slot->idx; + + if (add) { + g->mems_sts[idx/8] &= ~(1 << (idx%8)); + } + else { + g->mems_sts[idx/8] |= (1 << (idx%8)); + } + return 0; +} + static int piix4_device_hotplug(DeviceState *qdev, PCIDevice *dev, PCIHotplugState state) { diff --git a/hw/dimm.c b/hw/dimm.c index ba104cc..2115567 100644 --- a/hw/dimm.c +++ b/hw/dimm.c @@ -25,6 +25,7 @@ static DeviceState *dimm_hotplug_qdev; static dimm_hotplug_fn dimm_hotplug; +static dimm_hotplug_fn dimm_revert; static QTAILQ_HEAD(Dimmlist, DimmState) dimmlist; static QTAILQ_HEAD(dimm_hp_result_head, dimm_hp_result) dimm_hp_result_queue; @@ -77,10 +78,12 @@ DimmState *dimm_create(char *id, uint64_t size, uint64_t node, uint32_t return mdev; } -void dimm_register_hotplug(dimm_hotplug_fn hotplug, DeviceState *qdev) +void dimm_register_hotplug(dimm_hotplug_fn hotplug, dimm_hotplug_fn revert, + DeviceState *qdev) { dimm_hotplug_qdev = qdev; dimm_hotplug = hotplug; + dimm_revert = revert; dimm_scan_populated(); } @@ -211,10 +214,20 @@ void dimm_notify(uint32_t idx, uint32_t event) s->pending = false; break; case DIMM_REMOVE_FAIL: + QTAILQ_INSERT_TAIL(&dimm_hp_result_queue, result, next); + s->pending = false; + if (dimm_revert) + dimm_revert(dimm_hotplug_qdev, (SysBusDevice*)s, 0); + break; case DIMM_ADD_SUCCESS: + QTAILQ_INSERT_TAIL(&dimm_hp_result_queue, result, next); + s->pending = false; + break; case DIMM_ADD_FAIL: QTAILQ_INSERT_TAIL(&dimm_hp_result_queue, result, next); s->pending = false; + if (dimm_revert) + dimm_revert(dimm_hotplug_qdev, (SysBusDevice*)s, 1); break; default: g_free(result); @@ -288,6 +301,7 @@ static void dimm_class_init(ObjectClass *klass, void *data) dc->props = dimm_properties; sc->init = dimm_init; dimm_hotplug = NULL; + dimm_revert = NULL; QTAILQ_INIT(&dimmlist); QTAILQ_INIT(&dimm_hp_result_queue); } diff --git a/hw/dimm.h b/hw/dimm.h index 0fa6137..b563e3f 100644 --- a/hw/dimm.h +++ b/hw/dimm.h @@ -54,7 +54,7 @@ void dimm_depopulate(DimmState *s); int dimm_do(Monitor *mon, const QDict *qdict, bool add); DimmState *dimm_find_from_idx(uint32_t idx); DimmState *dimm_find_from_name(char *id); -void dimm_register_hotplug(dimm_hotplug_fn hotplug, DeviceState *qdev); +void dimm_register_hotplug(dimm_hotplug_fn hotplug, dimm_hotplug_fn revert, DeviceState *qdev); void dimm_calc_offsets(dimm_calcoffset_fn calcfn); void dimm_activate(DimmState *slot); void dimm_deactivate(DimmState *slot); -- 1.7.9

1 0

[RFC PATCH v2 17/21] acpi_dsdt: Revert internal dimm state on _OST failure
by Vasilis Liaskovitis 11 Jul '12

11 Jul '12

This reverts bitmap state in the case of a failed hot operation, in order to allow retry of failed hot operations Signed-off-by: Vasilis Liaskovitis <vasilis.liaskovitis(a)profitbricks.com> --- src/acpi-dsdt.dsl | 4 ++++ 1 files changed, 4 insertions(+), 0 deletions(-) diff --git a/src/acpi-dsdt.dsl b/src/acpi-dsdt.dsl index 1c253ca..0d37bbc 100644 --- a/src/acpi-dsdt.dsl +++ b/src/acpi-dsdt.dsl @@ -832,6 +832,8 @@ DefinitionBlock ( Switch(And(Arg1, 0xFF)) { Case(0x1) { Store(Arg2, MEF) + // Revert MEON flag for this memory device to one + Store(One, Index(MEON, Arg2)) } } } @@ -843,6 +845,8 @@ DefinitionBlock ( } Case(0x1) { Store(Arg2, MIF) + // Revert MEON flag for this memory device to zero + Store(Zero, Index(MEON, Arg2)) } } } -- 1.7.9

1 0

[RFC PATCH v2 16/21] acpi_piix4: Update dimm state on VM reboot
by Vasilis Liaskovitis 11 Jul '12

11 Jul '12

in case of hot-remove or hot-add failure, the dimm bitmaps in qemu and Seabios are inconsistent with the true state of the DIMM devices. The "populated" field of the DimmState reflects the true state of the device. This inconsistency means that a failed operation cannot be retried. Ths patch updates the bit array to the true state of the dimms on VM reboot. This allows retry of failed hot-add or hot-remove operations after a reboot. Retrying a failed hot operation is not yet possible before reboot (the following patch removes this limitation for guests with _OST acpi support) Signed-off-by: Vasilis Liaskovitis <vasilis.liaskovitis(a)profitbricks.com> --- hw/acpi_piix4.c | 25 +++++++++++++++++++++++++ 1 files changed, 25 insertions(+), 0 deletions(-) diff --git a/hw/acpi_piix4.c b/hw/acpi_piix4.c index d8e2c22..ebc5de7 100644 --- a/hw/acpi_piix4.c +++ b/hw/acpi_piix4.c @@ -91,6 +91,7 @@ typedef struct PIIX4PMState { } PIIX4PMState; static void piix4_acpi_system_hot_add_init(PCIBus *bus, PIIX4PMState *s); +static void piix4_dimm_state_sync(PIIX4PMState *s); #define ACPI_ENABLE 0xf1 #define ACPI_DISABLE 0xf0 @@ -369,6 +370,7 @@ static void piix4_reset(void *opaque) /* Mark SMM as already inited (until KVM supports SMM). */ pci_conf[0x5B] = 0x02; } + piix4_dimm_state_sync(s); piix4_update_hotplug(s); } @@ -671,6 +673,29 @@ static int piix4_dimm_hotplug(DeviceState *qdev, SysBusDevice *dev, int return 0; } +void piix4_dimm_state_sync(PIIX4PMState *s) +{ + struct gpe_regs *g = &s->gperegs; + DimmState *slot = NULL; + uint32_t i, temp = 1; + + for(i = 0; i < MAX_DIMMS; i++) { + slot = dimm_find_from_idx(i); + if (!slot) + break; + if (i % 8 == 0) { + temp = 1; + g->mems_sts[i / 8] = 0; + } + else + temp = temp << 1; + if (slot->populated) { + g->mems_sts[i / 8] |= temp; + } + slot->pending = false; + } +} + static int piix4_device_hotplug(DeviceState *qdev, PCIDevice *dev, PCIHotplugState state) { -- 1.7.9

1 0

[RFC PATCH v2 15/21] acpi_piix4: _OST dimm support
by Vasilis Liaskovitis 11 Jul '12

11 Jul '12

This allows qemu to receive notifications from the guest OS on success or failure of a memory hotplug request. The guest OS needs to implement the _OST functionality for this to work (linux-next: http://lkml.org/lkml/2012/6/25/321) Also add new _OST registers in docs/specs/acpi_hotplug.txt Signed-off-by: Vasilis Liaskovitis <vasilis.liaskovitis(a)profitbricks.com> --- docs/specs/acpi_hotplug.txt | 24 ++++++++++++++++++++++++ hw/acpi_piix4.c | 15 +++++++++++++++ hw/dimm.c | 18 ++++++++++++++++++ hw/dimm.h | 1 + 4 files changed, 58 insertions(+), 0 deletions(-) diff --git a/docs/specs/acpi_hotplug.txt b/docs/specs/acpi_hotplug.txt index cf86242..2f6fd5f 100644 --- a/docs/specs/acpi_hotplug.txt +++ b/docs/specs/acpi_hotplug.txt @@ -20,3 +20,27 @@ ejected. Written by ACPI memory device _EJ0 method to notify qemu of successfull hot-removal. Write-only. + +Memory Dimm ejection failure notification (IO port 0xafa1, 1-byte access): +--------------------------------------------------------------- +Dimm hot-remove _OST failure notification. Byte value indicates Dimm slot for +which ejection failed. + +Written by ACPI memory device _OST method to notify qemu of failed +hot-removal. Write-only. + +Memory Dimm insertion success notification (IO port 0xafa2, 1-byte access): +--------------------------------------------------------------- +Dimm hot-add _OST success notification. Byte value indicates Dimm slot for which +insertion succeeded. + +Written by ACPI memory device _OST method to notify qemu of failed +hot-add. Write-only. + +Memory Dimm insertion failure notification (IO port 0xafa3, 1-byte access): +--------------------------------------------------------------- +Dimm hot-add _OST failure notification. Byte value indicates Dimm slot for which +insertion failed. + +Written by ACPI memory device _OST method to notify qemu of failed +hot-add. Write-only. diff --git a/hw/acpi_piix4.c b/hw/acpi_piix4.c index b988597..d8e2c22 100644 --- a/hw/acpi_piix4.c +++ b/hw/acpi_piix4.c @@ -49,6 +49,9 @@ #define PCI_RMV_BASE 0xae0c #define MEM_BASE 0xaf80 #define MEM_EJ_BASE 0xafa0 +#define MEM_OST_REMOVE_FAIL 0xafa1 +#define MEM_OST_ADD_SUCCESS 0xafa2 +#define MEM_OST_ADD_FAIL 0xafa3 #define PIIX4_MEM_HOTPLUG_STATUS 8 #define PIIX4_PCI_HOTPLUG_STATUS 2 @@ -531,6 +534,15 @@ static void gpe_writeb(void *opaque, uint32_t addr, uint32_t val) case MEM_EJ_BASE: dimm_notify(val, DIMM_REMOVE_SUCCESS); break; + case MEM_OST_REMOVE_FAIL: + dimm_notify(val, DIMM_REMOVE_FAIL); + break; + case MEM_OST_ADD_SUCCESS: + dimm_notify(val, DIMM_ADD_SUCCESS); + break; + case MEM_OST_ADD_FAIL: + dimm_notify(val, DIMM_ADD_FAIL); + break; default: acpi_gpe_ioport_writeb(&s->ar, addr, val); } @@ -604,6 +616,9 @@ static void piix4_acpi_system_hot_add_init(PCIBus *bus, PIIX4PMState *s) register_ioport_read(MEM_BASE, DIMM_BITMAP_BYTES, 1, gpe_readb, s); register_ioport_write(MEM_EJ_BASE, 1, 1, gpe_writeb, s); + register_ioport_write(MEM_OST_REMOVE_FAIL, 1, 1, gpe_writeb, s); + register_ioport_write(MEM_OST_ADD_SUCCESS, 1, 1, gpe_writeb, s); + register_ioport_write(MEM_OST_ADD_FAIL, 1, 1, gpe_writeb, s); for(i = 0; i < DIMM_BITMAP_BYTES; i++) { s->gperegs.mems_sts[i] = 0; diff --git a/hw/dimm.c b/hw/dimm.c index 9b32386..ba104cc 100644 --- a/hw/dimm.c +++ b/hw/dimm.c @@ -89,12 +89,14 @@ void dimm_activate(DimmState *slot) dimm_populate(slot); if (dimm_hotplug) dimm_hotplug(dimm_hotplug_qdev, (SysBusDevice*)slot, 1); + slot->pending = true; } void dimm_deactivate(DimmState *slot) { if (dimm_hotplug) dimm_hotplug(dimm_hotplug_qdev, (SysBusDevice*)slot, 0); + slot->pending = true; } DimmState *dimm_find_from_name(char *id) @@ -138,6 +140,10 @@ int dimm_do(Monitor *mon, const QDict *qdict, bool add) __FUNCTION__, id); return 1; } + if (slot->pending) { + fprintf(stderr, "warning: %s slot %s hot-operation pending\n", + __FUNCTION__, id); + } dimm_activate(slot); } else { @@ -146,6 +152,10 @@ int dimm_do(Monitor *mon, const QDict *qdict, bool add) __FUNCTION__, id); return 1; } + if (slot->pending) { + fprintf(stderr, "warning: %s slot %s hot-operation pending\n", + __FUNCTION__, id); + } dimm_deactivate(slot); } @@ -198,6 +208,13 @@ void dimm_notify(uint32_t idx, uint32_t event) case DIMM_REMOVE_SUCCESS: dimm_depopulate(s); QTAILQ_INSERT_TAIL(&dimm_hp_result_queue, result, next); + s->pending = false; + break; + case DIMM_REMOVE_FAIL: + case DIMM_ADD_SUCCESS: + case DIMM_ADD_FAIL: + QTAILQ_INSERT_TAIL(&dimm_hp_result_queue, result, next); + s->pending = false; break; default: g_free(result); @@ -259,6 +276,7 @@ static int dimm_init(SysBusDevice *s) slot = DIMM(s); slot->mr = NULL; slot->populated = false; + slot->pending = false; return 0; } diff --git a/hw/dimm.h b/hw/dimm.h index 3e55ed3..0fa6137 100644 --- a/hw/dimm.h +++ b/hw/dimm.h @@ -35,6 +35,7 @@ typedef struct DimmState { MemoryRegion *mr; /* MemoryRegion for this slot. !NULL only if populated */ bool populated; /* 1 means device has been hotplugged. Default is 0. */ QTAILQ_ENTRY (DimmState) nextdimm; + bool pending; /* true means a hot operation is pending for this dimm */ } DimmState; struct dimm_hp_result { -- 1.7.9

1 0

[RFC PATCH v2 14/21] acpi_dsdt: Support _OST dimm method
by Vasilis Liaskovitis 11 Jul '12

11 Jul '12

Add support for _OST method. _OST method will write into the correct I/O byte to signal success / failure of hot-add or hot-remove to qemu. Signed-off-by: Vasilis Liaskovitis <vasilis.liaskovitis(a)profitbricks.com> --- src/acpi-dsdt.dsl | 46 ++++++++++++++++++++++++++++++++++++++++++++++ src/ssdt-mem.dsl | 4 ++++ 2 files changed, 50 insertions(+), 0 deletions(-) diff --git a/src/acpi-dsdt.dsl b/src/acpi-dsdt.dsl index 5d3e92b..1c253ca 100644 --- a/src/acpi-dsdt.dsl +++ b/src/acpi-dsdt.dsl @@ -762,6 +762,28 @@ DefinitionBlock ( MPE, 8 } + + /* Memory hot-remove notify failure byte */ + OperationRegion(MEEF, SystemIO, 0xafa1, 1) + Field (MEEF, ByteAcc, NoLock, Preserve) + { + MEF, 8 + } + + /* Memory hot-add notify success byte */ + OperationRegion(MPIS, SystemIO, 0xafa2, 1) + Field (MPIS, ByteAcc, NoLock, Preserve) + { + MIS, 8 + } + + /* Memory hot-add notify failure byte */ + OperationRegion(MPIF, SystemIO, 0xafa3, 1) + Field (MPIF, ByteAcc, NoLock, Preserve) + { + MIF, 8 + } + Method(MESC, 0) { // Local5 = active memdevice bitmap Store (MES, Local5) @@ -802,6 +824,30 @@ DefinitionBlock ( Store(Arg0, MPE) Sleep(200) } + Method (MOST, 3, Serialized) { + // _OST method - OS status indication + Switch (And(Arg0, 0xFF)) { + Case(0x3) + { + Switch(And(Arg1, 0xFF)) { + Case(0x1) { + Store(Arg2, MEF) + } + } + } + Case(0x1) + { + Switch(And(Arg1, 0xFF)) { + Case(0x0) { + Store(Arg2, MIS) + } + Case(0x1) { + Store(Arg2, MIF) + } + } + } + } + } } diff --git a/src/ssdt-mem.dsl b/src/ssdt-mem.dsl index ee322f0..041d301 100644 --- a/src/ssdt-mem.dsl +++ b/src/ssdt-mem.dsl @@ -38,6 +38,7 @@ DefinitionBlock ("ssdt-mem.aml", "SSDT", 0x02, "BXPC", "CSSDT", 0x1) External(CMST, MethodObj) External(MPEJ, MethodObj) + External(MOST, MethodObj) Name(_CRS, ResourceTemplate() { QwordMemory( @@ -60,6 +61,9 @@ DefinitionBlock ("ssdt-mem.aml", "SSDT", 0x02, "BXPC", "CSSDT", 0x1) Method (_EJ0, 1, NotSerialized) { MPEJ(ID, Arg0) } + Method (_OST, 3) { + MOST(Arg0, Arg1, ID) + } } } -- 1.7.9

1 0

[RFC PATCH v2 13/21] Implement memory hotplug notification lists
by Vasilis Liaskovitis 11 Jul '12

11 Jul '12

Guest can respond to ACPI hotplug events e.g. with _EJ or _OST method. This patch implements a tail queue to store guest notifications for memory hot-add and hot-remove requests. Guest responses for memory hotplug command on a per-dimm basis can be detected with the new hmp command "info memhp" or the new qmp command "query-memhp" Examples: (qemu) dimm_add dimm0 (qemu) info memhp Dimm: dimm0 hot-add success or Dimm: dimm0 hot-add failure (qemu) dimm_del dimm0 (qemu) info memhp Dimm: dimm0 hot-remove success or Dimm: dimm0 hot-remove failure Results are removed from the queue once read. This patch only queues _EJ events that signal hot-remove success. For _OST event queuing, which cover the hot-remove failure and hot-add success/failure cases, the next 2 patches are also needed. These notification items should probably be part of migration state (not yet implemented) Signed-off-by: Vasilis Liaskovitis <vasilis.liaskovitis(a)profitbricks.com> --- hmp-commands.hx | 2 + hmp.c | 17 ++++++++++++++++ hmp.h | 1 + hw/dimm.c | 55 ++++++++++++++++++++++++++++++++++++++++++++++++++++++ hw/dimm.h | 6 +++++ monitor.c | 7 ++++++ qapi-schema.json | 26 +++++++++++++++++++++++++ qmp-commands.hx | 38 +++++++++++++++++++++++++++++++++++++ 8 files changed, 152 insertions(+), 0 deletions(-) diff --git a/hmp-commands.hx b/hmp-commands.hx index 012c150..3172cde 100644 --- a/hmp-commands.hx +++ b/hmp-commands.hx @@ -1459,6 +1459,8 @@ show device tree show qdev device model list @item info roms show roms +@item info memhp +show memhp @end table ETEXI diff --git a/hmp.c b/hmp.c index b9cec1d..ec25d9a 100644 --- a/hmp.c +++ b/hmp.c @@ -1000,3 +1000,20 @@ void hmp_netdev_del(Monitor *mon, const QDict *qdict) qmp_netdev_del(id, &err); hmp_handle_error(mon, &err); } + +void hmp_info_memhp(Monitor *mon) +{ + MemHpInfoList *info; + MemHpInfoList *item; + MemHpInfo *dimm; + + info = qmp_query_memhp(NULL); + for (item = info; item; item = item->next) { + dimm = item->value; + monitor_printf(mon, "Dimm: %s %s %s\n", dimm->Dimm, + dimm->request, dimm->result); + dimm->Dimm = NULL; + } + + qapi_free_MemHpInfoList(info); +} diff --git a/hmp.h b/hmp.h index 79d138d..971e7c4 100644 --- a/hmp.h +++ b/hmp.h @@ -64,5 +64,6 @@ void hmp_device_del(Monitor *mon, const QDict *qdict); void hmp_dump_guest_memory(Monitor *mon, const QDict *qdict); void hmp_netdev_add(Monitor *mon, const QDict *qdict); void hmp_netdev_del(Monitor *mon, const QDict *qdict); +void hmp_info_memhp(Monitor *mon); #endif diff --git a/hw/dimm.c b/hw/dimm.c index 00c4623..9b32386 100644 --- a/hw/dimm.c +++ b/hw/dimm.c @@ -26,6 +26,7 @@ static DeviceState *dimm_hotplug_qdev; static dimm_hotplug_fn dimm_hotplug; static QTAILQ_HEAD(Dimmlist, DimmState) dimmlist; +static QTAILQ_HEAD(dimm_hp_result_head, dimm_hp_result) dimm_hp_result_queue; static Property dimm_properties[] = { DEFINE_PROP_END_OF_LIST() @@ -189,16 +190,69 @@ void dimm_notify(uint32_t idx, uint32_t event) DimmState *s; s = dimm_find_from_idx(idx); assert(s != NULL); + struct dimm_hp_result *result = g_malloc0(sizeof(*result)); + result->s = s; + result->ret = event; switch(event) { case DIMM_REMOVE_SUCCESS: dimm_depopulate(s); + QTAILQ_INSERT_TAIL(&dimm_hp_result_queue, result, next); break; default: + g_free(result); break; } } +MemHpInfoList *qmp_query_memhp(Error **errp) +{ + MemHpInfoList *head = NULL, *cur_item = NULL, *info; + struct dimm_hp_result *item, *nextitem; + + QTAILQ_FOREACH_SAFE(item, &dimm_hp_result_queue, next, nextitem) { + + info = g_malloc0(sizeof(*info)); + info->value = g_malloc0(sizeof(*info->value)); + info->value->Dimm = g_malloc0(sizeof(char) * 32); + info->value->request = g_malloc0(sizeof(char) * 16); + info->value->result = g_malloc0(sizeof(char) * 16); + switch (item->ret) { + case DIMM_REMOVE_SUCCESS: + strcpy(info->value->request, "hot-remove"); + strcpy(info->value->result, "success"); + break; + case DIMM_REMOVE_FAIL: + strcpy(info->value->request, "hot-remove"); + strcpy(info->value->result, "failure"); + break; + case DIMM_ADD_SUCCESS: + strcpy(info->value->request, "hot-add"); + strcpy(info->value->result, "success"); + break; + case DIMM_ADD_FAIL: + strcpy(info->value->request, "hot-add"); + strcpy(info->value->result, "failure"); + break; + default: + break; + } + strcpy(info->value->Dimm, item->s->busdev.qdev.id); + /* XXX: waiting for the qapi to support GSList */ + if (!cur_item) { + head = cur_item = info; + } else { + cur_item->next = info; + cur_item = info; + } + + /* hotplug notification copied to qmp list, delete original item */ + QTAILQ_REMOVE(&dimm_hp_result_queue, item, next); + g_free(item); + } + + return head; +} static int dimm_init(SysBusDevice *s) { DimmState *slot; @@ -217,6 +271,7 @@ static void dimm_class_init(ObjectClass *klass, void *data) sc->init = dimm_init; dimm_hotplug = NULL; QTAILQ_INIT(&dimmlist); + QTAILQ_INIT(&dimm_hp_result_queue); } static TypeInfo dimm_info = { diff --git a/hw/dimm.h b/hw/dimm.h index 643f319..3e55ed3 100644 --- a/hw/dimm.h +++ b/hw/dimm.h @@ -37,6 +37,12 @@ typedef struct DimmState { QTAILQ_ENTRY (DimmState) nextdimm; } DimmState; +struct dimm_hp_result { + DimmState *s; + dimm_hp_result_code ret; + QTAILQ_ENTRY (dimm_hp_result) next; +}; + typedef int (*dimm_hotplug_fn)(DeviceState *qdev, SysBusDevice *dev, int add); typedef target_phys_addr_t (*dimm_calcoffset_fn)(uint64_t size); diff --git a/monitor.c b/monitor.c index d3d95a6..4a14e26 100644 --- a/monitor.c +++ b/monitor.c @@ -2732,6 +2732,13 @@ static mon_cmd_t info_cmds[] = { .mhandler.info = do_trace_print_events, }, { + .name = "memhp", + .args_type = "", + .params = "", + .help = "show memory hotplug status", + .mhandler.info = hmp_info_memhp, + }, + { .name = NULL, }, }; diff --git a/qapi-schema.json b/qapi-schema.json index 3b6e346..049f6f9 100644 --- a/qapi-schema.json +++ b/qapi-schema.json @@ -1862,3 +1862,29 @@ # Since: 0.14.0 ## { 'command': 'netdev_del', 'data': {'id': 'str'} } + +## +# @MemHpInfo: +# +# Information about status of a memory hotplug command +# +# @Dimm: the Dimm associated with the result +# +# @result: the result of the hotplug command +# +# Since: 1.1.3 +# +## +{ 'type': 'MemHpInfo', + 'data': {'Dimm': 'str', 'request': 'str', 'result': 'str'} } + +## +# @query-memhp: +# +# Returns a list of information about pending hotplug commands +# +# Returns: a list of @MemhpInfo +# +# Since: 1.1.3 +## +{ 'command': 'query-memhp', 'returns': ['MemHpInfo'] } diff --git a/qmp-commands.hx b/qmp-commands.hx index 7efd628..cd1d5f0 100644 --- a/qmp-commands.hx +++ b/qmp-commands.hx @@ -2248,3 +2248,41 @@ Example: <- { "return": {} } EQMP + + { + .name = "query-memhp", + .args_type = "", + .mhandler.cmd_new = qmp_marshal_input_query_memhp + }, +SQMP +query-memhp +---------- + +Show memory hotplug command notifications. + +Return a json-array. Each DIMM that has a pending notification is represented +by a json-object, which contains: + +- "Dimm": Dimm name (json-str) +- "request": type of hot request: hot-add or hot-remove (json-str) +- "result": result of the hotplug request for this Dimm success or failure (json-str) + +Example: + +-> { "execute": "query-memhp" } +<- { + "return":[ + { + "result": "failure", + "request": "hot-remove", + "Dimm": "dimm10" + }, + { + "result": "success", + "request": "hot-add", + "Dimm": "dimm3" + } + ] + } + +EQMP -- 1.7.9

1 0

[RFC PATCH v2 12/21] fix live-migration when "populated=on" is missing
by Vasilis Liaskovitis 11 Jul '12

11 Jul '12

Live migration works after memory hot-add events, as long as the qemu command line "-dimm" arguments are changed on the destination host to specify "populated=on" for the dimms that have been hot-added. If a command-line change has not occured, the destination host does not yet have the corresponding ramblock in its ram_list. Activate the memslot on the destination during ram_load. Perhaps several fields of the DimmState struct should be part of a VMStateDescription to handle migration in a cleaner way. Signed-off-by: Vasilis Liaskovitis <vasilis.liaskovitis(a)profitbricks.com> --- arch_init.c | 23 ++++++++++++++++++++--- 1 files changed, 20 insertions(+), 3 deletions(-) diff --git a/arch_init.c b/arch_init.c index a9e8b74..5f46b98 100644 --- a/arch_init.c +++ b/arch_init.c @@ -43,6 +43,7 @@ #include "hw/smbios.h" #include "exec-memory.h" #include "hw/pcspk.h" +#include "hw/dimm.h" #ifdef TARGET_SPARC int graphic_width = 1024; @@ -452,9 +453,25 @@ int ram_load(QEMUFile *f, void *opaque, int version_id) } if (!block) { - fprintf(stderr, "Unknown ramblock \"%s\", cannot " - "accept migration\n", id); - return -EINVAL; + /* this can happen if a dimm was hot-added at source host */ + DimmState *slot = dimm_find_from_name(id); + if (slot) { + dimm_activate(slot); + /* rescan ram_list, verify ramblock is there now */ + QLIST_FOREACH(block, &ram_list.blocks, next) { + if (!strncmp(id, block->idstr, sizeof(id))) { + if (block->length != length) + return -EINVAL; + break; + } + } + assert(block); + } + else { + fprintf(stderr, "Unknown ramblock \"%s\", cannot " + "accept migration\n", id); + return -EINVAL; + } } total_ram_bytes -= length; -- 1.7.9

1 0

[RFC PATCH v2 11/21] Implement dimm_add and dimm_del hmp/qmp commands
by Vasilis Liaskovitis 11 Jul '12

11 Jul '12

Hot-add hmp syntax: dimm_add dimmid Hot-remove hmp syntax: dimm_del dimmid Respective qmp commands are "dimm-add", "dimm-del". Signed-off-by: Vasilis Liaskovitis <vasilis.liaskovitis(a)profitbricks.com> --- hmp-commands.hx | 32 ++++++++++++++++++++++++++++++++ monitor.c | 11 +++++++++++ monitor.h | 3 +++ qmp-commands.hx | 39 +++++++++++++++++++++++++++++++++++++++ 4 files changed, 85 insertions(+), 0 deletions(-) diff --git a/hmp-commands.hx b/hmp-commands.hx index f5d9d91..012c150 100644 --- a/hmp-commands.hx +++ b/hmp-commands.hx @@ -618,6 +618,38 @@ Add device. ETEXI { + .name = "dimm_del", + .args_type = "id:s", + .params = "id", + .help = "hot-remove memory (dimm device)", + .user_print = monitor_user_noop, + .mhandler.cmd_new = do_dimm_del, + }, + +STEXI +@item dimm_del @var{config} +@findex dimm_del + +Hot-remove dimm. +ETEXI + + { + .name = "dimm_add", + .args_type = "id:s", + .params = "id", + .help = "hot-add memory (dimm device)", + .user_print = monitor_user_noop, + .mhandler.cmd_new = do_dimm_add, + }, + +STEXI +@item dimm_add @var{config} +@findex dimm_add + +Hot-add dimm. +ETEXI + + { .name = "device_del", .args_type = "id:s", .params = "device", diff --git a/monitor.c b/monitor.c index f6107ba..d3d95a6 100644 --- a/monitor.c +++ b/monitor.c @@ -67,6 +67,7 @@ #include "qmp-commands.h" #include "hmp.h" #include "qemu-thread.h" +#include "hw/dimm.h" /* for pic/irq_info */ #if defined(TARGET_SPARC) @@ -4813,3 +4814,13 @@ int monitor_read_block_device_key(Monitor *mon, const char *device, return monitor_read_bdrv_key_start(mon, bs, completion_cb, opaque); } + +int do_dimm_add(Monitor *mon, const QDict *qdict, QObject **ret_data) +{ + return dimm_do(mon, qdict, true); +} + +int do_dimm_del(Monitor *mon, const QDict *qdict, QObject **ret_data) +{ + return dimm_do(mon, qdict, false); +} diff --git a/monitor.h b/monitor.h index 5f4de1b..afdd721 100644 --- a/monitor.h +++ b/monitor.h @@ -86,4 +86,7 @@ int qmp_qom_set(Monitor *mon, const QDict *qdict, QObject **ret); int qmp_qom_get(Monitor *mon, const QDict *qdict, QObject **ret); +int do_dimm_add(Monitor *mon, const QDict *qdict, QObject **ret_data); +int do_dimm_del(Monitor *mon, const QDict *qdict, QObject **ret_data); + #endif /* !MONITOR_H */ diff --git a/qmp-commands.hx b/qmp-commands.hx index 2e1a38e..7efd628 100644 --- a/qmp-commands.hx +++ b/qmp-commands.hx @@ -2209,3 +2209,42 @@ EQMP .args_type = "implements:s?,abstract:b?", .mhandler.cmd_new = qmp_marshal_input_qom_list_types, }, + { + .name = "dimm-add", + .args_type = "id:s", + .mhandler.cmd_new = do_dimm_add, + }, +SQMP +dimm-add +------------- + +Hot-add memory DIMM + +Will hotplug memory DIMMs with given id. + +Example: + +-> { "execute": "dimm-add", "arguments": { "id": "dimm0" } } +<- { "return": {} } + +EQMP + + { + .name = "dimm-del", + .args_type = "id:s", + .mhandler.cmd_new = do_dimm_del, + }, +SQMP +dimm-del +------------- + +Hot-remove memory DIMM + +Will hot-unplug memory DIMMs with given id. + +Example: + +-> { "execute": "dimm-del", "arguments": { "id": "dimm0" } } +<- { "return": {} } + +EQMP -- 1.7.9

1 0

[RFC PATCH v2 10/21] Implement "-dimm" command line option
by Vasilis Liaskovitis 11 Jul '12

11 Jul '12

Syntax: "-dimm id=name,size=sz,node=pxm,populated=on|off" The starting physical address for all dimms is calculated automatically from top of memory, skipping the pci hole at [PCI_HOLE_START, 4G). "populated=on" means the dimm is populated at machine startup. Default is off. "node" is defining numa proximity for this dimm. Default is node zero. Example: "-dimm id=dimm0,size=512M,node=0,populated=off" will define a 512M memory slot belonging to numa node 0. Signed-off-by: Vasilis Liaskovitis <vasilis.liaskovitis(a)profitbricks.com> --- qemu-config.c | 25 +++++++++++++++++++++++++ qemu-options.hx | 5 +++++ sysemu.h | 1 + vl.c | 35 +++++++++++++++++++++++++++++++++++ 4 files changed, 66 insertions(+), 0 deletions(-) diff --git a/qemu-config.c b/qemu-config.c index 5c3296b..4abc31b 100644 --- a/qemu-config.c +++ b/qemu-config.c @@ -626,6 +626,30 @@ QemuOptsList qemu_boot_opts = { }, }; +static QemuOptsList qemu_dimm_opts = { + .name = "dimm", + .head = QTAILQ_HEAD_INITIALIZER(qemu_dimm_opts.head), + .desc = { + { + .name = "id", + .type = QEMU_OPT_STRING, + .help = "id of this dimm device", + },{ + .name = "size", + .type = QEMU_OPT_SIZE, + .help = "memory size for this dimm", + },{ + .name = "populated", + .type = QEMU_OPT_BOOL, + .help = "populated for this dimm", + },{ + .name = "node", + .type = QEMU_OPT_NUMBER, + .help = "NUMA node number (i.e. proximity) for this dimm", + }, + { /* end of list */ } + }, +}; static QemuOptsList *vm_config_groups[32] = { &qemu_drive_opts, &qemu_chardev_opts, @@ -641,6 +665,7 @@ static QemuOptsList *vm_config_groups[32] = { &qemu_machine_opts, &qemu_boot_opts, &qemu_iscsi_opts, + &qemu_dimm_opts, NULL, }; diff --git a/qemu-options.hx b/qemu-options.hx index 8b66264..61909f7 100644 --- a/qemu-options.hx +++ b/qemu-options.hx @@ -2747,3 +2747,8 @@ HXCOMM This is the last statement. Insert new options before this line! STEXI @end table ETEXI + +DEF("dimm", HAS_ARG, QEMU_OPTION_dimm, + "-dimm id=dimmid,size=sz,node=nd,populated=on|off\n" + "specify memory dimm device with name dimmid, size sz on node nd", + QEMU_ARCH_ALL) diff --git a/sysemu.h b/sysemu.h index bc2c788..3e21a22 100644 --- a/sysemu.h +++ b/sysemu.h @@ -136,6 +136,7 @@ extern QEMUClock *rtc_clock; extern int nb_numa_nodes; extern uint64_t node_mem[MAX_NODES]; extern uint64_t node_cpumask[MAX_NODES]; +extern int nb_hp_dimms; #define MAX_OPTION_ROMS 16 typedef struct QEMUOptionRom { diff --git a/vl.c b/vl.c index 0ff8818..efe915e 100644 --- a/vl.c +++ b/vl.c @@ -120,6 +120,7 @@ int main(int argc, char **argv) #include "hw/xen.h" #include "hw/qdev.h" #include "hw/loader.h" +#include "hw/dimm.h" #include "bt-host.h" #include "net.h" #include "net/slirp.h" @@ -242,6 +243,7 @@ QTAILQ_HEAD(, FWBootEntry) fw_boot_order = QTAILQ_HEAD_INITIALIZER(fw_boot_order int nb_numa_nodes; uint64_t node_mem[MAX_NODES]; uint64_t node_cpumask[MAX_NODES]; +int nb_hp_dimms; uint8_t qemu_uuid[16]; @@ -518,6 +520,23 @@ static void configure_rtc_date_offset(const char *startdate, int legacy) rtc_date_offset = time(NULL) - rtc_start_date; } } +static void configure_dimm(QemuOpts *opts) +{ + const char *id; + uint64_t size, node; + bool populated; + if (nb_hp_dimms == MAX_DIMMS) { + fprintf(stderr, "qemu: maximum number of DIMMs (%d) exceeded\n", + MAX_DIMMS); + exit(1); + } + id = qemu_opts_id(opts); + size = qemu_opt_get_size(opts, "size", DEFAULT_DIMMSIZE); + populated = qemu_opt_get_bool(opts, "populated", 0); + node = qemu_opt_get_number(opts, "node", 0); + dimm_create((char*)id, size, node, nb_hp_dimms, populated); + nb_hp_dimms++; +} static void configure_rtc(QemuOpts *opts) { @@ -2273,6 +2292,8 @@ int main(int argc, char **argv, char **envp) DisplayChangeListener *dcl; int cyls, heads, secs, translation; QemuOpts *hda_opts = NULL, *opts, *machine_opts; + QemuOpts *dimm_opts[MAX_DIMMS]; + int nb_dimm_opts = 0; QemuOptsList *olist; int optind; const char *optarg; @@ -3200,6 +3221,18 @@ int main(int argc, char **argv, char **envp) case QEMU_OPTION_qtest_log: qtest_log = optarg; break; + case QEMU_OPTION_dimm: + if (nb_dimm_opts == MAX_DIMMS) { + fprintf(stderr, "qemu: maximum number of DIMMs (%d) exceeded\n", + MAX_DIMMS); + } + dimm_opts[nb_dimm_opts] = + qemu_opts_parse(qemu_find_opts("dimm"), optarg, 0); + if (!dimm_opts[nb_dimm_opts]) { + exit(1); + } + nb_dimm_opts++; + break; default: os_parse_cmd_args(popt->index, optarg); } @@ -3517,6 +3550,8 @@ int main(int argc, char **argv, char **envp) } qemu_add_globals(); + for (i = 0; i < nb_dimm_opts; i++) + configure_dimm(dimm_opts[i]); qdev_machine_init(); machine->init(ram_size, boot_devices, -- 1.7.9

1 0

[RFC PATCH v2 09/21] pc: Add dimm paravirt SRAT info
by Vasilis Liaskovitis 11 Jul '12

11 Jul '12

The numa_fw_cfg paravirt interface is extended to include SRAT information for all hotplug-able dimms. There are 3 words for each hotplug-able memory slot, denoting start address, size and node proximity. The new info is appended after existing numa info, so that the fw_cfg layout does not break. This information is used by Seabios to build hotplug memory device objects at runtime. nb_numa_nodes is set to 1 by default (not 0), so that we always pass srat info to SeaBIOS. v1->v2: Dimm SRAT info (#dimms) is appended at end of existing numa fw_cfg in order not to break existing layout Documentation of the new fwcfg layout is included in docs/specs/fwcfg.txt Signed-off-by: Vasilis Liaskovitis <vasilis.liaskovitis(a)profitbricks.com> --- docs/specs/fwcfg.txt | 28 ++++++++++++++++++++++++++ hw/pc.c | 53 ++++++++++++++++++++++++++++++++++++++++++++++++- vl.c | 2 +- 3 files changed, 80 insertions(+), 3 deletions(-) create mode 100644 docs/specs/fwcfg.txt diff --git a/docs/specs/fwcfg.txt b/docs/specs/fwcfg.txt new file mode 100644 index 0000000..e6fcd8f --- /dev/null +++ b/docs/specs/fwcfg.txt @@ -0,0 +1,28 @@ +QEMU<->BIOS Paravirt Documentation +-------------------------------------- + +This document describes paravirt data structures passed from QEMU to BIOS. + +fw_cfg SRAT paravirt info +-------------------- +The SRAT info passed from QEMU to BIOS has the following layout: + +----------------------------------------------------------------------------------------------- +#nodes | cpu0_pxm | cpu1_pxm | ... | cpulast_pxm | node0_mem | node1_mem | ... | nodelast_mem + +----------------------------------------------------------------------------------------------- +#dimms | dimm0_start | dimm0_sz | dimm0_pxm | ... | dimmlast_start | dimmlast_sz | dimmlast_pxm + +Entry 0 contains the number of numa nodes (nb_numa_nodes). + +Entries 1..max_cpus: The next max_cpus entries describe node proximity for each +one of the vCPUs in the system. + +Entries max_cpus+1..max_cpus+nb_numa_nodes+1: The next nb_numa_nodes entries +describe the memory size for each one of the NUMA nodes in the system. + +Entry max_cpus+nb_numa_nodes+1 contains the number of memory dimms (nb_hp_dimms) + +The last 3 * nb_hp_dimms entries are organized in triplets: Each triplet contains +the physical address offset, size (in bytes), and node proximity for the +respective dimm. diff --git a/hw/pc.c b/hw/pc.c index ef9901a..cf651d0 100644 --- a/hw/pc.c +++ b/hw/pc.c @@ -598,12 +598,15 @@ int e820_add_entry(uint64_t address, uint64_t length, uint32_t type) return index; } +static void setup_hp_dimms(uint64_t *fw_cfg_slots); + static void *bochs_bios_init(void) { void *fw_cfg; uint8_t *smbios_table; size_t smbios_len; uint64_t *numa_fw_cfg; + uint64_t *hp_dimms_fw_cfg; int i, j; register_ioport_write(0x400, 1, 2, bochs_bios_write, NULL); @@ -638,8 +641,10 @@ static void *bochs_bios_init(void) /* allocate memory for the NUMA channel: one (64bit) word for the number * of nodes, one word for each VCPU->node and one word for each node to * hold the amount of memory. + * Finally one word for the number of hotplug memory slots and three words + * for each hotplug memory slot (start address, size and node proximity). */ - numa_fw_cfg = g_malloc0((1 + max_cpus + nb_numa_nodes) * 8); + numa_fw_cfg = g_malloc0((2 + max_cpus + nb_numa_nodes + 3 * nb_hp_dimms) * 8); numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes); for (i = 0; i < max_cpus; i++) { for (j = 0; j < nb_numa_nodes; j++) { @@ -652,8 +657,15 @@ static void *bochs_bios_init(void) for (i = 0; i < nb_numa_nodes; i++) { numa_fw_cfg[max_cpus + 1 + i] = cpu_to_le64(node_mem[i]); } + + numa_fw_cfg[1 + max_cpus + nb_numa_nodes] = cpu_to_le64(nb_hp_dimms); + + hp_dimms_fw_cfg = numa_fw_cfg + 2 + max_cpus + nb_numa_nodes; + if (nb_hp_dimms) + setup_hp_dimms(hp_dimms_fw_cfg); + fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, (uint8_t *)numa_fw_cfg, - (1 + max_cpus + nb_numa_nodes) * 8); + (2 + max_cpus + nb_numa_nodes + 3 * nb_hp_dimms) * 8); return fw_cfg; } @@ -1223,3 +1235,40 @@ target_phys_addr_t pc_set_hp_memory_offset(uint64_t size) return ret; } + +static void setup_hp_dimms(uint64_t *fw_cfg_slots) +{ + int i = 0; + Error *err = NULL; + DeviceState *dev; + DimmState *slot; + const char *type; + BusChild *kid; + BusState *bus = sysbus_get_default(); + + QTAILQ_FOREACH(kid, &bus->children, sibling) { + dev = kid->child; + type = object_property_get_str(OBJECT(dev), "type", &err); + if (err) { + error_free(err); + fprintf(stderr, "error getting device type\n"); + exit(1); + } + + if (!strcmp(type, "dimm")) { + if (!dev->id) { + fprintf(stderr, "error getting dimm device id\n"); + exit(1); + } + slot = DIMM(dev); + /* determine starting physical address for this memory slot */ + assert(slot->start); + fw_cfg_slots[3 * slot->idx] = cpu_to_le64(slot->start); + fw_cfg_slots[3 * slot->idx + 1] = cpu_to_le64(slot->size); + fw_cfg_slots[3 * slot->idx + 2] = cpu_to_le64(slot->node); + i++; + } + } + assert(i == nb_hp_dimms); +} + diff --git a/vl.c b/vl.c index 0ff8818..37c9798 100644 --- a/vl.c +++ b/vl.c @@ -2335,7 +2335,7 @@ int main(int argc, char **argv, char **envp) node_cpumask[i] = 0; } - nb_numa_nodes = 0; + nb_numa_nodes = 1; nb_nics = 0; autostart= 1; -- 1.7.9

1 0

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

SeaBIOS July 2012