qemu/hw/ppc/spapr_drc.c

875 lines
30 KiB
C
Raw Normal View History

/*
* QEMU SPAPR Dynamic Reconfiguration Connector Implementation
*
* Copyright IBM Corp. 2014
*
* Authors:
* Michael Roth <mdroth@linux.vnet.ibm.com>
*
* This work is licensed under the terms of the GNU GPL, version 2 or later.
* See the COPYING file in the top-level directory.
*/
#include "qemu/osdep.h"
2016-03-14 11:01:28 +03:00
#include "qapi/error.h"
#include "cpu.h"
#include "qemu/cutils.h"
#include "hw/ppc/spapr_drc.h"
#include "qom/object.h"
#include "hw/qdev.h"
#include "qapi/visitor.h"
#include "qemu/error-report.h"
#include "hw/ppc/spapr.h" /* for RTAS return codes */
#include "trace.h"
#define DRC_CONTAINER_PATH "/dr-connector"
#define DRC_INDEX_TYPE_SHIFT 28
#define DRC_INDEX_ID_MASK ((1ULL << DRC_INDEX_TYPE_SHIFT) - 1)
static sPAPRDRConnectorTypeShift get_type_shift(sPAPRDRConnectorType type)
{
uint32_t shift = 0;
/* make sure this isn't SPAPR_DR_CONNECTOR_TYPE_ANY, or some
* other wonky value.
*/
g_assert(is_power_of_2(type));
while (type != (1 << shift)) {
shift++;
}
return shift;
}
static uint32_t get_index(sPAPRDRConnector *drc)
{
/* no set format for a drc index: it only needs to be globally
* unique. this is how we encode the DRC type on bare-metal
* however, so might as well do that here
*/
return (get_type_shift(drc->type) << DRC_INDEX_TYPE_SHIFT) |
(drc->id & DRC_INDEX_ID_MASK);
}
static uint32_t set_isolation_state(sPAPRDRConnector *drc,
sPAPRDRIsolationState state)
{
sPAPRDRConnectorClass *drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
trace_spapr_drc_set_isolation_state(get_index(drc), state);
if (state == SPAPR_DR_ISOLATION_STATE_UNISOLATED) {
/* cannot unisolate a non-existent resource, and, or resources
* which are in an 'UNUSABLE' allocation state. (PAPR 2.7, 13.5.3.5)
*/
if (!drc->dev ||
drc->allocation_state == SPAPR_DR_ALLOCATION_STATE_UNUSABLE) {
return RTAS_OUT_NO_SUCH_INDICATOR;
}
}
/*
* Fail any requests to ISOLATE the LMB DRC if this LMB doesn't
* belong to a DIMM device that is marked for removal.
*
* Currently the guest userspace tool drmgr that drives the memory
* hotplug/unplug will just try to remove a set of 'removable' LMBs
* in response to a hot unplug request that is based on drc-count.
* If the LMB being removed doesn't belong to a DIMM device that is
* actually being unplugged, fail the isolation request here.
*/
if (drc->type == SPAPR_DR_CONNECTOR_TYPE_LMB) {
if ((state == SPAPR_DR_ISOLATION_STATE_ISOLATED) &&
!drc->awaiting_release) {
return RTAS_OUT_HW_ERROR;
}
}
drc->isolation_state = state;
if (drc->isolation_state == SPAPR_DR_ISOLATION_STATE_ISOLATED) {
/* if we're awaiting release, but still in an unconfigured state,
* it's likely the guest is still in the process of configuring
* the device and is transitioning the devices to an ISOLATED
* state as a part of that process. so we only complete the
* removal when this transition happens for a device in a
* configured state, as suggested by the state diagram from
* PAPR+ 2.7, 13.4
*/
if (drc->awaiting_release) {
if (drc->configured) {
trace_spapr_drc_set_isolation_state_finalizing(get_index(drc));
drck->detach(drc, DEVICE(drc->dev), drc->detach_cb,
drc->detach_cb_opaque, NULL);
} else {
trace_spapr_drc_set_isolation_state_deferring(get_index(drc));
}
}
drc->configured = false;
}
return RTAS_OUT_SUCCESS;
}
static uint32_t set_indicator_state(sPAPRDRConnector *drc,
sPAPRDRIndicatorState state)
{
trace_spapr_drc_set_indicator_state(get_index(drc), state);
drc->indicator_state = state;
return RTAS_OUT_SUCCESS;
}
static uint32_t set_allocation_state(sPAPRDRConnector *drc,
sPAPRDRAllocationState state)
{
sPAPRDRConnectorClass *drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
trace_spapr_drc_set_allocation_state(get_index(drc), state);
if (state == SPAPR_DR_ALLOCATION_STATE_USABLE) {
/* if there's no resource/device associated with the DRC, there's
* no way for us to put it in an allocation state consistent with
* being 'USABLE'. PAPR 2.7, 13.5.3.4 documents that this should
* result in an RTAS return code of -3 / "no such indicator"
*/
if (!drc->dev) {
return RTAS_OUT_NO_SUCH_INDICATOR;
}
2017-03-28 15:09:34 +03:00
if (drc->awaiting_release && drc->awaiting_allocation) {
/* kernel is acknowledging a previous hotplug event
* while we are already removing it.
* it's safe to ignore awaiting_allocation here since we know the
* situation is predicated on the guest either already having done
* so (boot-time hotplug), or never being able to acquire in the
* first place (hotplug followed by immediate unplug).
*/
drc->awaiting_allocation_skippable = true;
return RTAS_OUT_NO_SUCH_INDICATOR;
}
}
if (drc->type != SPAPR_DR_CONNECTOR_TYPE_PCI) {
drc->allocation_state = state;
if (drc->awaiting_release &&
drc->allocation_state == SPAPR_DR_ALLOCATION_STATE_UNUSABLE) {
trace_spapr_drc_set_allocation_state_finalizing(get_index(drc));
drck->detach(drc, DEVICE(drc->dev), drc->detach_cb,
drc->detach_cb_opaque, NULL);
} else if (drc->allocation_state == SPAPR_DR_ALLOCATION_STATE_USABLE) {
drc->awaiting_allocation = false;
}
}
return RTAS_OUT_SUCCESS;
}
static uint32_t get_type(sPAPRDRConnector *drc)
{
return drc->type;
}
static const char *get_name(sPAPRDRConnector *drc)
{
return drc->name;
}
static const void *get_fdt(sPAPRDRConnector *drc, int *fdt_start_offset)
{
if (fdt_start_offset) {
*fdt_start_offset = drc->fdt_start_offset;
}
return drc->fdt;
}
static void set_configured(sPAPRDRConnector *drc)
{
trace_spapr_drc_set_configured(get_index(drc));
if (drc->isolation_state != SPAPR_DR_ISOLATION_STATE_UNISOLATED) {
/* guest should be not configuring an isolated device */
trace_spapr_drc_set_configured_skipping(get_index(drc));
return;
}
drc->configured = true;
}
spapr_drc: enable immediate detach for unsignalled devices Currently spapr doesn't support "aborting" hotplug of PCI devices by allowing device_del to immediately remove the device if we haven't signalled the presence of the device to the guest. In the past this wasn't an issue, since we always immediately signalled device attach and simply relied on full guest-aware add->remove path for device removal. However, as of 788d259, we now defer signalling for PCI functions until function 0 is attached, so now we need to deal with these "abort" operations for cases where a user hotplugs a non-0 function, then opts to remove it prior hotplugging function 0. Currently they'd have to reboot before the unplug completed. PCIe multifunction hotplug does not have this requirement however, so from a management implementation perspective it would be good to address this within the same release as 788d259. We accomplish this by simply adding a 'signalled' flag to track whether a device hotplug event has been sent to the guest. If it hasn't, we allow immediate removal under the assumption that the guest will not be using the device. Devices present at boot/reset time are also assumed to be 'signalled'. For CPU/memory/etc, signalling will still happen immediately as part of device_add, so only PCI functions should be affected. Cc: bharata@linux.vnet.ibm.com Cc: david@gibson.dropbear.id.au Cc: sbhat@linux.vnet.ibm.com Cc: qemu-ppc@nongnu.org Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com> [dwg: This fixes a regression where an incorrect hot-add of a non-zero function can no longer be backed out until function 0 is added] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2016-04-01 01:27:37 +03:00
/* has the guest been notified of device attachment? */
static void set_signalled(sPAPRDRConnector *drc)
{
drc->signalled = true;
}
/*
* dr-entity-sense sensor value
* returned via get-sensor-state RTAS calls
* as expected by state diagram in PAPR+ 2.7, 13.4
* based on the current allocation/indicator/power states
* for the DR connector.
*/
static uint32_t entity_sense(sPAPRDRConnector *drc, sPAPRDREntitySense *state)
{
if (drc->dev) {
if (drc->type != SPAPR_DR_CONNECTOR_TYPE_PCI &&
drc->allocation_state == SPAPR_DR_ALLOCATION_STATE_UNUSABLE) {
/* for logical DR, we return a state of UNUSABLE
* iff the allocation state UNUSABLE.
* Otherwise, report the state as USABLE/PRESENT,
* as we would for PCI.
*/
*state = SPAPR_DR_ENTITY_SENSE_UNUSABLE;
} else {
/* this assumes all PCI devices are assigned to
* a 'live insertion' power domain, where QEMU
* manages power state automatically as opposed
* to the guest. present, non-PCI resources are
* unaffected by power state.
*/
*state = SPAPR_DR_ENTITY_SENSE_PRESENT;
}
} else {
if (drc->type == SPAPR_DR_CONNECTOR_TYPE_PCI) {
/* PCI devices, and only PCI devices, use EMPTY
* in cases where we'd otherwise use UNUSABLE
*/
*state = SPAPR_DR_ENTITY_SENSE_EMPTY;
} else {
*state = SPAPR_DR_ENTITY_SENSE_UNUSABLE;
}
}
trace_spapr_drc_entity_sense(get_index(drc), *state);
return RTAS_OUT_SUCCESS;
}
static void prop_get_index(Object *obj, Visitor *v, const char *name,
void *opaque, Error **errp)
{
sPAPRDRConnector *drc = SPAPR_DR_CONNECTOR(obj);
sPAPRDRConnectorClass *drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
uint32_t value = (uint32_t)drck->get_index(drc);
qapi: Swap visit_* arguments for consistent 'name' placement JSON uses "name":value, but many of our visitor interfaces were called with visit_type_FOO(v, &value, name, errp). This can be a bit confusing to have to mentally swap the parameter order to match JSON order. It's particularly bad for visit_start_struct(), where the 'name' parameter is smack in the middle of the otherwise-related group of 'obj, kind, size' parameters! It's time to do a global swap of the parameter ordering, so that the 'name' parameter is always immediately after the Visitor argument. Additional reason in favor of the swap: the existing include/qjson.h prefers listing 'name' first in json_prop_*(), and I have plans to unify that file with the qapi visitors; listing 'name' first in qapi will minimize churn to the (admittedly few) qjson.h clients. Later patches will then fix docs, object.h, visitor-impl.h, and those clients to match. Done by first patching scripts/qapi*.py by hand to make generated files do what I want, then by running the following Coccinelle script to affect the rest of the code base: $ spatch --sp-file script `git grep -l '\bvisit_' -- '**/*.[ch]'` I then had to apply some touchups (Coccinelle insisted on TAB indentation in visitor.h, and botched the signature of visit_type_enum() by rewriting 'const char *const strings[]' to the syntactically invalid 'const char*const[] strings'). The movement of parameters is sufficient to provoke compiler errors if any callers were missed. // Part 1: Swap declaration order @@ type TV, TErr, TObj, T1, T2; identifier OBJ, ARG1, ARG2; @@ void visit_start_struct -(TV v, TObj OBJ, T1 ARG1, const char *name, T2 ARG2, TErr errp) +(TV v, const char *name, TObj OBJ, T1 ARG1, T2 ARG2, TErr errp) { ... } @@ type bool, TV, T1; identifier ARG1; @@ bool visit_optional -(TV v, T1 ARG1, const char *name) +(TV v, const char *name, T1 ARG1) { ... } @@ type TV, TErr, TObj, T1; identifier OBJ, ARG1; @@ void visit_get_next_type -(TV v, TObj OBJ, T1 ARG1, const char *name, TErr errp) +(TV v, const char *name, TObj OBJ, T1 ARG1, TErr errp) { ... } @@ type TV, TErr, TObj, T1, T2; identifier OBJ, ARG1, ARG2; @@ void visit_type_enum -(TV v, TObj OBJ, T1 ARG1, T2 ARG2, const char *name, TErr errp) +(TV v, const char *name, TObj OBJ, T1 ARG1, T2 ARG2, TErr errp) { ... } @@ type TV, TErr, TObj; identifier OBJ; identifier VISIT_TYPE =~ "^visit_type_"; @@ void VISIT_TYPE -(TV v, TObj OBJ, const char *name, TErr errp) +(TV v, const char *name, TObj OBJ, TErr errp) { ... } // Part 2: swap caller order @@ expression V, NAME, OBJ, ARG1, ARG2, ERR; identifier VISIT_TYPE =~ "^visit_type_"; @@ ( -visit_start_struct(V, OBJ, ARG1, NAME, ARG2, ERR) +visit_start_struct(V, NAME, OBJ, ARG1, ARG2, ERR) | -visit_optional(V, ARG1, NAME) +visit_optional(V, NAME, ARG1) | -visit_get_next_type(V, OBJ, ARG1, NAME, ERR) +visit_get_next_type(V, NAME, OBJ, ARG1, ERR) | -visit_type_enum(V, OBJ, ARG1, ARG2, NAME, ERR) +visit_type_enum(V, NAME, OBJ, ARG1, ARG2, ERR) | -VISIT_TYPE(V, OBJ, NAME, ERR) +VISIT_TYPE(V, NAME, OBJ, ERR) ) Signed-off-by: Eric Blake <eblake@redhat.com> Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com> Message-Id: <1454075341-13658-19-git-send-email-eblake@redhat.com> Signed-off-by: Markus Armbruster <armbru@redhat.com>
2016-01-29 16:48:54 +03:00
visit_type_uint32(v, name, &value, errp);
}
static void prop_get_type(Object *obj, Visitor *v, const char *name,
void *opaque, Error **errp)
{
sPAPRDRConnector *drc = SPAPR_DR_CONNECTOR(obj);
sPAPRDRConnectorClass *drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
uint32_t value = (uint32_t)drck->get_type(drc);
qapi: Swap visit_* arguments for consistent 'name' placement JSON uses "name":value, but many of our visitor interfaces were called with visit_type_FOO(v, &value, name, errp). This can be a bit confusing to have to mentally swap the parameter order to match JSON order. It's particularly bad for visit_start_struct(), where the 'name' parameter is smack in the middle of the otherwise-related group of 'obj, kind, size' parameters! It's time to do a global swap of the parameter ordering, so that the 'name' parameter is always immediately after the Visitor argument. Additional reason in favor of the swap: the existing include/qjson.h prefers listing 'name' first in json_prop_*(), and I have plans to unify that file with the qapi visitors; listing 'name' first in qapi will minimize churn to the (admittedly few) qjson.h clients. Later patches will then fix docs, object.h, visitor-impl.h, and those clients to match. Done by first patching scripts/qapi*.py by hand to make generated files do what I want, then by running the following Coccinelle script to affect the rest of the code base: $ spatch --sp-file script `git grep -l '\bvisit_' -- '**/*.[ch]'` I then had to apply some touchups (Coccinelle insisted on TAB indentation in visitor.h, and botched the signature of visit_type_enum() by rewriting 'const char *const strings[]' to the syntactically invalid 'const char*const[] strings'). The movement of parameters is sufficient to provoke compiler errors if any callers were missed. // Part 1: Swap declaration order @@ type TV, TErr, TObj, T1, T2; identifier OBJ, ARG1, ARG2; @@ void visit_start_struct -(TV v, TObj OBJ, T1 ARG1, const char *name, T2 ARG2, TErr errp) +(TV v, const char *name, TObj OBJ, T1 ARG1, T2 ARG2, TErr errp) { ... } @@ type bool, TV, T1; identifier ARG1; @@ bool visit_optional -(TV v, T1 ARG1, const char *name) +(TV v, const char *name, T1 ARG1) { ... } @@ type TV, TErr, TObj, T1; identifier OBJ, ARG1; @@ void visit_get_next_type -(TV v, TObj OBJ, T1 ARG1, const char *name, TErr errp) +(TV v, const char *name, TObj OBJ, T1 ARG1, TErr errp) { ... } @@ type TV, TErr, TObj, T1, T2; identifier OBJ, ARG1, ARG2; @@ void visit_type_enum -(TV v, TObj OBJ, T1 ARG1, T2 ARG2, const char *name, TErr errp) +(TV v, const char *name, TObj OBJ, T1 ARG1, T2 ARG2, TErr errp) { ... } @@ type TV, TErr, TObj; identifier OBJ; identifier VISIT_TYPE =~ "^visit_type_"; @@ void VISIT_TYPE -(TV v, TObj OBJ, const char *name, TErr errp) +(TV v, const char *name, TObj OBJ, TErr errp) { ... } // Part 2: swap caller order @@ expression V, NAME, OBJ, ARG1, ARG2, ERR; identifier VISIT_TYPE =~ "^visit_type_"; @@ ( -visit_start_struct(V, OBJ, ARG1, NAME, ARG2, ERR) +visit_start_struct(V, NAME, OBJ, ARG1, ARG2, ERR) | -visit_optional(V, ARG1, NAME) +visit_optional(V, NAME, ARG1) | -visit_get_next_type(V, OBJ, ARG1, NAME, ERR) +visit_get_next_type(V, NAME, OBJ, ARG1, ERR) | -visit_type_enum(V, OBJ, ARG1, ARG2, NAME, ERR) +visit_type_enum(V, NAME, OBJ, ARG1, ARG2, ERR) | -VISIT_TYPE(V, OBJ, NAME, ERR) +VISIT_TYPE(V, NAME, OBJ, ERR) ) Signed-off-by: Eric Blake <eblake@redhat.com> Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com> Message-Id: <1454075341-13658-19-git-send-email-eblake@redhat.com> Signed-off-by: Markus Armbruster <armbru@redhat.com>
2016-01-29 16:48:54 +03:00
visit_type_uint32(v, name, &value, errp);
}
static char *prop_get_name(Object *obj, Error **errp)
{
sPAPRDRConnector *drc = SPAPR_DR_CONNECTOR(obj);
sPAPRDRConnectorClass *drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
return g_strdup(drck->get_name(drc));
}
static void prop_get_entity_sense(Object *obj, Visitor *v, const char *name,
void *opaque, Error **errp)
{
sPAPRDRConnector *drc = SPAPR_DR_CONNECTOR(obj);
sPAPRDRConnectorClass *drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
uint32_t value;
drck->entity_sense(drc, &value);
qapi: Swap visit_* arguments for consistent 'name' placement JSON uses "name":value, but many of our visitor interfaces were called with visit_type_FOO(v, &value, name, errp). This can be a bit confusing to have to mentally swap the parameter order to match JSON order. It's particularly bad for visit_start_struct(), where the 'name' parameter is smack in the middle of the otherwise-related group of 'obj, kind, size' parameters! It's time to do a global swap of the parameter ordering, so that the 'name' parameter is always immediately after the Visitor argument. Additional reason in favor of the swap: the existing include/qjson.h prefers listing 'name' first in json_prop_*(), and I have plans to unify that file with the qapi visitors; listing 'name' first in qapi will minimize churn to the (admittedly few) qjson.h clients. Later patches will then fix docs, object.h, visitor-impl.h, and those clients to match. Done by first patching scripts/qapi*.py by hand to make generated files do what I want, then by running the following Coccinelle script to affect the rest of the code base: $ spatch --sp-file script `git grep -l '\bvisit_' -- '**/*.[ch]'` I then had to apply some touchups (Coccinelle insisted on TAB indentation in visitor.h, and botched the signature of visit_type_enum() by rewriting 'const char *const strings[]' to the syntactically invalid 'const char*const[] strings'). The movement of parameters is sufficient to provoke compiler errors if any callers were missed. // Part 1: Swap declaration order @@ type TV, TErr, TObj, T1, T2; identifier OBJ, ARG1, ARG2; @@ void visit_start_struct -(TV v, TObj OBJ, T1 ARG1, const char *name, T2 ARG2, TErr errp) +(TV v, const char *name, TObj OBJ, T1 ARG1, T2 ARG2, TErr errp) { ... } @@ type bool, TV, T1; identifier ARG1; @@ bool visit_optional -(TV v, T1 ARG1, const char *name) +(TV v, const char *name, T1 ARG1) { ... } @@ type TV, TErr, TObj, T1; identifier OBJ, ARG1; @@ void visit_get_next_type -(TV v, TObj OBJ, T1 ARG1, const char *name, TErr errp) +(TV v, const char *name, TObj OBJ, T1 ARG1, TErr errp) { ... } @@ type TV, TErr, TObj, T1, T2; identifier OBJ, ARG1, ARG2; @@ void visit_type_enum -(TV v, TObj OBJ, T1 ARG1, T2 ARG2, const char *name, TErr errp) +(TV v, const char *name, TObj OBJ, T1 ARG1, T2 ARG2, TErr errp) { ... } @@ type TV, TErr, TObj; identifier OBJ; identifier VISIT_TYPE =~ "^visit_type_"; @@ void VISIT_TYPE -(TV v, TObj OBJ, const char *name, TErr errp) +(TV v, const char *name, TObj OBJ, TErr errp) { ... } // Part 2: swap caller order @@ expression V, NAME, OBJ, ARG1, ARG2, ERR; identifier VISIT_TYPE =~ "^visit_type_"; @@ ( -visit_start_struct(V, OBJ, ARG1, NAME, ARG2, ERR) +visit_start_struct(V, NAME, OBJ, ARG1, ARG2, ERR) | -visit_optional(V, ARG1, NAME) +visit_optional(V, NAME, ARG1) | -visit_get_next_type(V, OBJ, ARG1, NAME, ERR) +visit_get_next_type(V, NAME, OBJ, ARG1, ERR) | -visit_type_enum(V, OBJ, ARG1, ARG2, NAME, ERR) +visit_type_enum(V, NAME, OBJ, ARG1, ARG2, ERR) | -VISIT_TYPE(V, OBJ, NAME, ERR) +VISIT_TYPE(V, NAME, OBJ, ERR) ) Signed-off-by: Eric Blake <eblake@redhat.com> Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com> Message-Id: <1454075341-13658-19-git-send-email-eblake@redhat.com> Signed-off-by: Markus Armbruster <armbru@redhat.com>
2016-01-29 16:48:54 +03:00
visit_type_uint32(v, name, &value, errp);
}
static void prop_get_fdt(Object *obj, Visitor *v, const char *name,
void *opaque, Error **errp)
{
sPAPRDRConnector *drc = SPAPR_DR_CONNECTOR(obj);
Error *err = NULL;
int fdt_offset_next, fdt_offset, fdt_depth;
void *fdt;
if (!drc->fdt) {
visit_type_null(v, NULL, errp);
return;
}
fdt = drc->fdt;
fdt_offset = drc->fdt_start_offset;
fdt_depth = 0;
do {
const char *name = NULL;
const struct fdt_property *prop = NULL;
int prop_len = 0, name_len = 0;
uint32_t tag;
tag = fdt_next_tag(fdt, fdt_offset, &fdt_offset_next);
switch (tag) {
case FDT_BEGIN_NODE:
fdt_depth++;
name = fdt_get_name(fdt, fdt_offset, &name_len);
visit_start_struct(v, name, NULL, 0, &err);
if (err) {
error_propagate(errp, err);
return;
}
break;
case FDT_END_NODE:
/* shouldn't ever see an FDT_END_NODE before FDT_BEGIN_NODE */
g_assert(fdt_depth > 0);
qapi: Split visit_end_struct() into pieces As mentioned in previous patches, we want to call visit_end_struct() functions unconditionally, so that visitors can release resources tied up since the matching visit_start_struct() without also having to worry about error priority if more than one error occurs. Even though error_propagate() can be safely used to ignore a second error during cleanup caused by a first error, it is simpler if the cleanup cannot set an error. So, split out the error checking portion (basically, input visitors checking for unvisited keys) into a new function visit_check_struct(), which can be safely skipped if any earlier errors are encountered, and leave the cleanup portion (which never fails, but must be called unconditionally if visit_start_struct() succeeded) in visit_end_struct(). Generated code in qapi-visit.c has diffs resembling: |@@ -59,10 +59,12 @@ void visit_type_ACPIOSTInfo(Visitor *v, | goto out_obj; | } | visit_type_ACPIOSTInfo_members(v, obj, &err); |- error_propagate(errp, err); |- err = NULL; |+ if (err) { |+ goto out_obj; |+ } |+ visit_check_struct(v, &err); | out_obj: |- visit_end_struct(v, &err); |+ visit_end_struct(v); | out: and in qapi-event.c: @@ -47,7 +47,10 @@ void qapi_event_send_acpi_device_ost(ACP | goto out; | } | visit_type_q_obj_ACPI_DEVICE_OST_arg_members(v, &param, &err); |- visit_end_struct(v, err ? NULL : &err); |+ if (!err) { |+ visit_check_struct(v, &err); |+ } |+ visit_end_struct(v); | if (err) { | goto out; Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <1461879932-9020-20-git-send-email-eblake@redhat.com> [Conflict with a doc fixup resolved] Signed-off-by: Markus Armbruster <armbru@redhat.com>
2016-04-29 00:45:27 +03:00
visit_check_struct(v, &err);
2016-06-09 19:48:34 +03:00
visit_end_struct(v, NULL);
if (err) {
error_propagate(errp, err);
return;
}
fdt_depth--;
break;
case FDT_PROP: {
int i;
prop = fdt_get_property_by_offset(fdt, fdt_offset, &prop_len);
name = fdt_string(fdt, fdt32_to_cpu(prop->nameoff));
qapi: Simplify semantics of visit_next_list() The semantics of the list visit are somewhat baroque, with the following pseudocode when FooList is used: start() for (prev = head; cur = next(prev); prev = &cur) { visit(&cur->value) } Note that these semantics (advance before visit) requires that the first call to next() return the list head, while all other calls return the next element of the list; that is, every visitor implementation is required to track extra state to decide whether to return the input as-is, or to advance. It also requires an argument of 'GenericList **' to next(), solely because the first iteration might need to modify the caller's GenericList head, so that all other calls have to do a layer of dereferencing. Thankfully, we only have two uses of list visits in the entire code base: one in spapr_drc (which completely avoids visit_next_list(), feeding in integers from a different source than uint8List), and one in qapi-visit.py. That is, all other list visitors are generated in qapi-visit.c, and share the same paradigm based on a qapi FooList type, so we can refactor how lists are laid out with minimal churn among clients. We can greatly simplify things by hoisting the special case into the start() routine, and flipping the order in the loop to visit before advance: start(head) for (tail = *head; tail; tail = next(tail)) { visit(&tail->value) } With the simpler semantics, visitors have less state to track, the argument to next() is reduced to 'GenericList *', and it also becomes obvious whether an input visitor is allocating a FooList during visit_start_list() (rather than the old way of not knowing if an allocation happened until the first visit_next_list()). As a minor drawback, we now allocate in two functions instead of one, and have to pass the size to both functions (unless we were to tweak the input visitors to cache the size to start_list for reuse during next_list, but that defeats the goal of less visitor state). The signature of visit_start_list() is chosen to match visit_start_struct(), with the new parameters after 'name'. The spapr_drc case is a virtual visit, done by passing NULL for list, similarly to how NULL is passed to visit_start_struct() when a qapi type is not used in those visits. It was easy to provide these semantics for qmp-output and dealloc visitors, and a bit harder for qmp-input (several prerequisite patches refactored things to make this patch straightforward). But it turned out that the string and opts visitors munge enough other state during visit_next_list() to make it easier to just document and require a GenericList visit for now; an assertion will remind us to adjust things if we need the semantics in the future. Several pre-requisite cleanup patches made the reshuffling of the various visitors easier; particularly the qmp input visitor. Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <1461879932-9020-24-git-send-email-eblake@redhat.com> Signed-off-by: Markus Armbruster <armbru@redhat.com>
2016-04-29 00:45:31 +03:00
visit_start_list(v, name, NULL, 0, &err);
if (err) {
error_propagate(errp, err);
return;
}
for (i = 0; i < prop_len; i++) {
qapi: Swap visit_* arguments for consistent 'name' placement JSON uses "name":value, but many of our visitor interfaces were called with visit_type_FOO(v, &value, name, errp). This can be a bit confusing to have to mentally swap the parameter order to match JSON order. It's particularly bad for visit_start_struct(), where the 'name' parameter is smack in the middle of the otherwise-related group of 'obj, kind, size' parameters! It's time to do a global swap of the parameter ordering, so that the 'name' parameter is always immediately after the Visitor argument. Additional reason in favor of the swap: the existing include/qjson.h prefers listing 'name' first in json_prop_*(), and I have plans to unify that file with the qapi visitors; listing 'name' first in qapi will minimize churn to the (admittedly few) qjson.h clients. Later patches will then fix docs, object.h, visitor-impl.h, and those clients to match. Done by first patching scripts/qapi*.py by hand to make generated files do what I want, then by running the following Coccinelle script to affect the rest of the code base: $ spatch --sp-file script `git grep -l '\bvisit_' -- '**/*.[ch]'` I then had to apply some touchups (Coccinelle insisted on TAB indentation in visitor.h, and botched the signature of visit_type_enum() by rewriting 'const char *const strings[]' to the syntactically invalid 'const char*const[] strings'). The movement of parameters is sufficient to provoke compiler errors if any callers were missed. // Part 1: Swap declaration order @@ type TV, TErr, TObj, T1, T2; identifier OBJ, ARG1, ARG2; @@ void visit_start_struct -(TV v, TObj OBJ, T1 ARG1, const char *name, T2 ARG2, TErr errp) +(TV v, const char *name, TObj OBJ, T1 ARG1, T2 ARG2, TErr errp) { ... } @@ type bool, TV, T1; identifier ARG1; @@ bool visit_optional -(TV v, T1 ARG1, const char *name) +(TV v, const char *name, T1 ARG1) { ... } @@ type TV, TErr, TObj, T1; identifier OBJ, ARG1; @@ void visit_get_next_type -(TV v, TObj OBJ, T1 ARG1, const char *name, TErr errp) +(TV v, const char *name, TObj OBJ, T1 ARG1, TErr errp) { ... } @@ type TV, TErr, TObj, T1, T2; identifier OBJ, ARG1, ARG2; @@ void visit_type_enum -(TV v, TObj OBJ, T1 ARG1, T2 ARG2, const char *name, TErr errp) +(TV v, const char *name, TObj OBJ, T1 ARG1, T2 ARG2, TErr errp) { ... } @@ type TV, TErr, TObj; identifier OBJ; identifier VISIT_TYPE =~ "^visit_type_"; @@ void VISIT_TYPE -(TV v, TObj OBJ, const char *name, TErr errp) +(TV v, const char *name, TObj OBJ, TErr errp) { ... } // Part 2: swap caller order @@ expression V, NAME, OBJ, ARG1, ARG2, ERR; identifier VISIT_TYPE =~ "^visit_type_"; @@ ( -visit_start_struct(V, OBJ, ARG1, NAME, ARG2, ERR) +visit_start_struct(V, NAME, OBJ, ARG1, ARG2, ERR) | -visit_optional(V, ARG1, NAME) +visit_optional(V, NAME, ARG1) | -visit_get_next_type(V, OBJ, ARG1, NAME, ERR) +visit_get_next_type(V, NAME, OBJ, ARG1, ERR) | -visit_type_enum(V, OBJ, ARG1, ARG2, NAME, ERR) +visit_type_enum(V, NAME, OBJ, ARG1, ARG2, ERR) | -VISIT_TYPE(V, OBJ, NAME, ERR) +VISIT_TYPE(V, NAME, OBJ, ERR) ) Signed-off-by: Eric Blake <eblake@redhat.com> Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com> Message-Id: <1454075341-13658-19-git-send-email-eblake@redhat.com> Signed-off-by: Markus Armbruster <armbru@redhat.com>
2016-01-29 16:48:54 +03:00
visit_type_uint8(v, NULL, (uint8_t *)&prop->data[i], &err);
if (err) {
error_propagate(errp, err);
return;
}
}
visit_check_list(v, &err);
2016-06-09 19:48:34 +03:00
visit_end_list(v, NULL);
if (err) {
error_propagate(errp, err);
return;
}
break;
}
default:
error_setg(&error_abort, "device FDT in unexpected state: %d", tag);
}
fdt_offset = fdt_offset_next;
} while (fdt_depth != 0);
}
static void attach(sPAPRDRConnector *drc, DeviceState *d, void *fdt,
int fdt_start_offset, bool coldplug, Error **errp)
{
trace_spapr_drc_attach(get_index(drc));
if (drc->isolation_state != SPAPR_DR_ISOLATION_STATE_ISOLATED) {
error_setg(errp, "an attached device is still awaiting release");
return;
}
if (drc->type == SPAPR_DR_CONNECTOR_TYPE_PCI) {
g_assert(drc->allocation_state == SPAPR_DR_ALLOCATION_STATE_USABLE);
}
g_assert(fdt || coldplug);
/* NOTE: setting initial isolation state to UNISOLATED means we can't
* detach unless guest has a userspace/kernel that moves this state
* back to ISOLATED in response to an unplug event, or this is done
* manually by the admin prior. if we force things while the guest
* may be accessing the device, we can easily crash the guest, so we
* we defer completion of removal in such cases to the reset() hook.
*/
if (drc->type == SPAPR_DR_CONNECTOR_TYPE_PCI) {
drc->isolation_state = SPAPR_DR_ISOLATION_STATE_UNISOLATED;
}
drc->indicator_state = SPAPR_DR_INDICATOR_STATE_ACTIVE;
drc->dev = d;
drc->fdt = fdt;
drc->fdt_start_offset = fdt_start_offset;
drc->configured = coldplug;
/* 'logical' DR resources such as memory/cpus are in some cases treated
* as a pool of resources from which the guest is free to choose from
* based on only a count. for resources that can be assigned in this
* fashion, we must assume the resource is signalled immediately
* since a single hotplug request might make an arbitrary number of
* such attached resources available to the guest, as opposed to
* 'physical' DR resources such as PCI where each device/resource is
* signalled individually.
*/
drc->signalled = (drc->type != SPAPR_DR_CONNECTOR_TYPE_PCI)
? true : coldplug;
if (drc->type != SPAPR_DR_CONNECTOR_TYPE_PCI) {
drc->awaiting_allocation = true;
}
object_property_add_link(OBJECT(drc), "device",
object_get_typename(OBJECT(drc->dev)),
(Object **)(&drc->dev),
NULL, 0, NULL);
}
static void detach(sPAPRDRConnector *drc, DeviceState *d,
spapr_drc_detach_cb *detach_cb,
void *detach_cb_opaque, Error **errp)
{
trace_spapr_drc_detach(get_index(drc));
drc->detach_cb = detach_cb;
drc->detach_cb_opaque = detach_cb_opaque;
spapr_drc: enable immediate detach for unsignalled devices Currently spapr doesn't support "aborting" hotplug of PCI devices by allowing device_del to immediately remove the device if we haven't signalled the presence of the device to the guest. In the past this wasn't an issue, since we always immediately signalled device attach and simply relied on full guest-aware add->remove path for device removal. However, as of 788d259, we now defer signalling for PCI functions until function 0 is attached, so now we need to deal with these "abort" operations for cases where a user hotplugs a non-0 function, then opts to remove it prior hotplugging function 0. Currently they'd have to reboot before the unplug completed. PCIe multifunction hotplug does not have this requirement however, so from a management implementation perspective it would be good to address this within the same release as 788d259. We accomplish this by simply adding a 'signalled' flag to track whether a device hotplug event has been sent to the guest. If it hasn't, we allow immediate removal under the assumption that the guest will not be using the device. Devices present at boot/reset time are also assumed to be 'signalled'. For CPU/memory/etc, signalling will still happen immediately as part of device_add, so only PCI functions should be affected. Cc: bharata@linux.vnet.ibm.com Cc: david@gibson.dropbear.id.au Cc: sbhat@linux.vnet.ibm.com Cc: qemu-ppc@nongnu.org Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com> [dwg: This fixes a regression where an incorrect hot-add of a non-zero function can no longer be backed out until function 0 is added] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2016-04-01 01:27:37 +03:00
/* if we've signalled device presence to the guest, or if the guest
* has gone ahead and configured the device (via manually-executed
* device add via drmgr in guest, namely), we need to wait
* for the guest to quiesce the device before completing detach.
* Otherwise, we can assume the guest hasn't seen it and complete the
* detach immediately. Note that there is a small race window
* just before, or during, configuration, which is this context
* refers mainly to fetching the device tree via RTAS.
* During this window the device access will be arbitrated by
* associated DRC, which will simply fail the RTAS calls as invalid.
* This is recoverable within guest and current implementations of
* drmgr should be able to cope.
*/
if (!drc->signalled && !drc->configured) {
/* if the guest hasn't seen the device we can't rely on it to
* set it back to an isolated state via RTAS, so do it here manually
*/
drc->isolation_state = SPAPR_DR_ISOLATION_STATE_ISOLATED;
}
if (drc->isolation_state != SPAPR_DR_ISOLATION_STATE_ISOLATED) {
trace_spapr_drc_awaiting_isolated(get_index(drc));
drc->awaiting_release = true;
return;
}
if (drc->type != SPAPR_DR_CONNECTOR_TYPE_PCI &&
drc->allocation_state != SPAPR_DR_ALLOCATION_STATE_UNUSABLE) {
trace_spapr_drc_awaiting_unusable(get_index(drc));
drc->awaiting_release = true;
return;
}
if (drc->awaiting_allocation) {
2017-03-28 15:09:34 +03:00
if (!drc->awaiting_allocation_skippable) {
drc->awaiting_release = true;
trace_spapr_drc_awaiting_allocation(get_index(drc));
return;
}
}
drc->indicator_state = SPAPR_DR_INDICATOR_STATE_INACTIVE;
if (drc->detach_cb) {
drc->detach_cb(drc->dev, drc->detach_cb_opaque);
}
drc->awaiting_release = false;
2017-03-28 15:09:34 +03:00
drc->awaiting_allocation_skippable = false;
g_free(drc->fdt);
drc->fdt = NULL;
drc->fdt_start_offset = 0;
object_property_del(OBJECT(drc), "device", NULL);
drc->dev = NULL;
drc->detach_cb = NULL;
drc->detach_cb_opaque = NULL;
}
static bool release_pending(sPAPRDRConnector *drc)
{
return drc->awaiting_release;
}
static void reset(DeviceState *d)
{
sPAPRDRConnector *drc = SPAPR_DR_CONNECTOR(d);
sPAPRDRConnectorClass *drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
spapr_drc: enable immediate detach for unsignalled devices Currently spapr doesn't support "aborting" hotplug of PCI devices by allowing device_del to immediately remove the device if we haven't signalled the presence of the device to the guest. In the past this wasn't an issue, since we always immediately signalled device attach and simply relied on full guest-aware add->remove path for device removal. However, as of 788d259, we now defer signalling for PCI functions until function 0 is attached, so now we need to deal with these "abort" operations for cases where a user hotplugs a non-0 function, then opts to remove it prior hotplugging function 0. Currently they'd have to reboot before the unplug completed. PCIe multifunction hotplug does not have this requirement however, so from a management implementation perspective it would be good to address this within the same release as 788d259. We accomplish this by simply adding a 'signalled' flag to track whether a device hotplug event has been sent to the guest. If it hasn't, we allow immediate removal under the assumption that the guest will not be using the device. Devices present at boot/reset time are also assumed to be 'signalled'. For CPU/memory/etc, signalling will still happen immediately as part of device_add, so only PCI functions should be affected. Cc: bharata@linux.vnet.ibm.com Cc: david@gibson.dropbear.id.au Cc: sbhat@linux.vnet.ibm.com Cc: qemu-ppc@nongnu.org Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com> [dwg: This fixes a regression where an incorrect hot-add of a non-zero function can no longer be backed out until function 0 is added] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2016-04-01 01:27:37 +03:00
sPAPRDREntitySense state;
trace_spapr_drc_reset(drck->get_index(drc));
/* immediately upon reset we can safely assume DRCs whose devices
* are pending removal can be safely removed, and that they will
* subsequently be left in an ISOLATED state. move the DRC to this
* state in these cases (which will in turn complete any pending
* device removals)
*/
if (drc->awaiting_release) {
drck->set_isolation_state(drc, SPAPR_DR_ISOLATION_STATE_ISOLATED);
/* generally this should also finalize the removal, but if the device
* hasn't yet been configured we normally defer removal under the
* assumption that this transition is taking place as part of device
* configuration. so check if we're still waiting after this, and
* force removal if we are
*/
if (drc->awaiting_release) {
drck->detach(drc, DEVICE(drc->dev), drc->detach_cb,
drc->detach_cb_opaque, NULL);
}
/* non-PCI devices may be awaiting a transition to UNUSABLE */
if (drc->type != SPAPR_DR_CONNECTOR_TYPE_PCI &&
drc->awaiting_release) {
drck->set_allocation_state(drc, SPAPR_DR_ALLOCATION_STATE_UNUSABLE);
}
}
spapr_drc: enable immediate detach for unsignalled devices Currently spapr doesn't support "aborting" hotplug of PCI devices by allowing device_del to immediately remove the device if we haven't signalled the presence of the device to the guest. In the past this wasn't an issue, since we always immediately signalled device attach and simply relied on full guest-aware add->remove path for device removal. However, as of 788d259, we now defer signalling for PCI functions until function 0 is attached, so now we need to deal with these "abort" operations for cases where a user hotplugs a non-0 function, then opts to remove it prior hotplugging function 0. Currently they'd have to reboot before the unplug completed. PCIe multifunction hotplug does not have this requirement however, so from a management implementation perspective it would be good to address this within the same release as 788d259. We accomplish this by simply adding a 'signalled' flag to track whether a device hotplug event has been sent to the guest. If it hasn't, we allow immediate removal under the assumption that the guest will not be using the device. Devices present at boot/reset time are also assumed to be 'signalled'. For CPU/memory/etc, signalling will still happen immediately as part of device_add, so only PCI functions should be affected. Cc: bharata@linux.vnet.ibm.com Cc: david@gibson.dropbear.id.au Cc: sbhat@linux.vnet.ibm.com Cc: qemu-ppc@nongnu.org Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com> [dwg: This fixes a regression where an incorrect hot-add of a non-zero function can no longer be backed out until function 0 is added] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2016-04-01 01:27:37 +03:00
drck->entity_sense(drc, &state);
if (state == SPAPR_DR_ENTITY_SENSE_PRESENT) {
drck->set_signalled(drc);
}
}
static void realize(DeviceState *d, Error **errp)
{
sPAPRDRConnector *drc = SPAPR_DR_CONNECTOR(d);
sPAPRDRConnectorClass *drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
Object *root_container;
char link_name[256];
gchar *child_name;
Error *err = NULL;
trace_spapr_drc_realize(drck->get_index(drc));
/* NOTE: we do this as part of realize/unrealize due to the fact
* that the guest will communicate with the DRC via RTAS calls
* referencing the global DRC index. By unlinking the DRC
* from DRC_CONTAINER_PATH/<drc_index> we effectively make it
* inaccessible by the guest, since lookups rely on this path
* existing in the composition tree
*/
root_container = container_get(object_get_root(), DRC_CONTAINER_PATH);
snprintf(link_name, sizeof(link_name), "%x", drck->get_index(drc));
child_name = object_get_canonical_path_component(OBJECT(drc));
trace_spapr_drc_realize_child(drck->get_index(drc), child_name);
object_property_add_alias(root_container, link_name,
drc->owner, child_name, &err);
if (err) {
error_report_err(err);
object_unref(OBJECT(drc));
}
g_free(child_name);
trace_spapr_drc_realize_complete(drck->get_index(drc));
}
static void unrealize(DeviceState *d, Error **errp)
{
sPAPRDRConnector *drc = SPAPR_DR_CONNECTOR(d);
sPAPRDRConnectorClass *drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
Object *root_container;
char name[256];
Error *err = NULL;
trace_spapr_drc_unrealize(drck->get_index(drc));
root_container = container_get(object_get_root(), DRC_CONTAINER_PATH);
snprintf(name, sizeof(name), "%x", drck->get_index(drc));
object_property_del(root_container, name, &err);
if (err) {
error_report_err(err);
object_unref(OBJECT(drc));
}
}
sPAPRDRConnector *spapr_dr_connector_new(Object *owner,
sPAPRDRConnectorType type,
uint32_t id)
{
sPAPRDRConnector *drc =
SPAPR_DR_CONNECTOR(object_new(TYPE_SPAPR_DR_CONNECTOR));
spapr: Don't use QOM [*] syntax for DR connectors. The dynamic reconfiguration (hotplug) code for the pseries machine type uses a "DR connector" QOM object for each resource it will be possible to hotplug. Each of these is added to its owner using object_property_add_child(owner, "dr-connector[*], ...); That works ok, mostly, but it means that the property indices are arbitrary, depending on the order in which the connectors are constructed. That might line up to something useful, but it doesn't have to. It will get worse once we add hotplug RAM support. That will add a DR connector object for every 256MB of potential memory. So if maxmem=2T, for example, there are 8192 objects under the same parent. The QOM interfaces aren't really designed for this. In particular object_property_add() with [*] has O(n^2) time complexity (in the number of existing children): first it has a linear search through array indices to find a free slot, each of which is attempted to a recursive call to object_property_add() with a specific [N]. Those calls are O(n) because there's a linear search through all properties to check for duplicates. By using a meaningful index value, which we already know is unique we can avoid the [*] special behaviour. That lets us reduce the total time for creating the DR objects from O(n^3) to O(n^2). O(n^2) is still kind of crappy, but it's enough to reduce the startup time of qemu (with in-progress memory hotplug support) with maxmem=2T from ~20 minutes to ~4 seconds. Signed-off-by: David Gibson <david@gibson.dropbear.id.au> Cc: Bharata B Rao <bharata@linux.vnet.ibm.com> Tested-by: Bharata B Rao <bharata@linux.vnet.ibm.com> Reviewed-by: Alexey Kardashevskiy <aik@ozlabs.ru>
2015-09-16 09:57:51 +03:00
char *prop_name;
g_assert(type);
drc->type = type;
drc->id = id;
drc->owner = owner;
spapr: Don't use QOM [*] syntax for DR connectors. The dynamic reconfiguration (hotplug) code for the pseries machine type uses a "DR connector" QOM object for each resource it will be possible to hotplug. Each of these is added to its owner using object_property_add_child(owner, "dr-connector[*], ...); That works ok, mostly, but it means that the property indices are arbitrary, depending on the order in which the connectors are constructed. That might line up to something useful, but it doesn't have to. It will get worse once we add hotplug RAM support. That will add a DR connector object for every 256MB of potential memory. So if maxmem=2T, for example, there are 8192 objects under the same parent. The QOM interfaces aren't really designed for this. In particular object_property_add() with [*] has O(n^2) time complexity (in the number of existing children): first it has a linear search through array indices to find a free slot, each of which is attempted to a recursive call to object_property_add() with a specific [N]. Those calls are O(n) because there's a linear search through all properties to check for duplicates. By using a meaningful index value, which we already know is unique we can avoid the [*] special behaviour. That lets us reduce the total time for creating the DR objects from O(n^3) to O(n^2). O(n^2) is still kind of crappy, but it's enough to reduce the startup time of qemu (with in-progress memory hotplug support) with maxmem=2T from ~20 minutes to ~4 seconds. Signed-off-by: David Gibson <david@gibson.dropbear.id.au> Cc: Bharata B Rao <bharata@linux.vnet.ibm.com> Tested-by: Bharata B Rao <bharata@linux.vnet.ibm.com> Reviewed-by: Alexey Kardashevskiy <aik@ozlabs.ru>
2015-09-16 09:57:51 +03:00
prop_name = g_strdup_printf("dr-connector[%"PRIu32"]", get_index(drc));
object_property_add_child(owner, prop_name, OBJECT(drc), NULL);
object_property_set_bool(OBJECT(drc), true, "realized", NULL);
spapr: Don't use QOM [*] syntax for DR connectors. The dynamic reconfiguration (hotplug) code for the pseries machine type uses a "DR connector" QOM object for each resource it will be possible to hotplug. Each of these is added to its owner using object_property_add_child(owner, "dr-connector[*], ...); That works ok, mostly, but it means that the property indices are arbitrary, depending on the order in which the connectors are constructed. That might line up to something useful, but it doesn't have to. It will get worse once we add hotplug RAM support. That will add a DR connector object for every 256MB of potential memory. So if maxmem=2T, for example, there are 8192 objects under the same parent. The QOM interfaces aren't really designed for this. In particular object_property_add() with [*] has O(n^2) time complexity (in the number of existing children): first it has a linear search through array indices to find a free slot, each of which is attempted to a recursive call to object_property_add() with a specific [N]. Those calls are O(n) because there's a linear search through all properties to check for duplicates. By using a meaningful index value, which we already know is unique we can avoid the [*] special behaviour. That lets us reduce the total time for creating the DR objects from O(n^3) to O(n^2). O(n^2) is still kind of crappy, but it's enough to reduce the startup time of qemu (with in-progress memory hotplug support) with maxmem=2T from ~20 minutes to ~4 seconds. Signed-off-by: David Gibson <david@gibson.dropbear.id.au> Cc: Bharata B Rao <bharata@linux.vnet.ibm.com> Tested-by: Bharata B Rao <bharata@linux.vnet.ibm.com> Reviewed-by: Alexey Kardashevskiy <aik@ozlabs.ru>
2015-09-16 09:57:51 +03:00
g_free(prop_name);
/* human-readable name for a DRC to encode into the DT
* description. this is mainly only used within a guest in place
* of the unique DRC index.
*
* in the case of VIO/PCI devices, it corresponds to a
* "location code" that maps a logical device/function (DRC index)
* to a physical (or virtual in the case of VIO) location in the
* system by chaining together the "location label" for each
* encapsulating component.
*
* since this is more to do with diagnosing physical hardware
* issues than guest compatibility, we choose location codes/DRC
* names that adhere to the documented format, but avoid encoding
* the entire topology information into the label/code, instead
* just using the location codes based on the labels for the
* endpoints (VIO/PCI adaptor connectors), which is basically
* just "C" followed by an integer ID.
*
* DRC names as documented by PAPR+ v2.7, 13.5.2.4
* location codes as documented by PAPR+ v2.7, 12.3.1.5
*/
switch (drc->type) {
case SPAPR_DR_CONNECTOR_TYPE_CPU:
drc->name = g_strdup_printf("CPU %d", id);
break;
case SPAPR_DR_CONNECTOR_TYPE_PHB:
drc->name = g_strdup_printf("PHB %d", id);
break;
case SPAPR_DR_CONNECTOR_TYPE_VIO:
case SPAPR_DR_CONNECTOR_TYPE_PCI:
drc->name = g_strdup_printf("C%d", id);
break;
case SPAPR_DR_CONNECTOR_TYPE_LMB:
drc->name = g_strdup_printf("LMB %d", id);
break;
default:
g_assert(false);
}
/* PCI slot always start in a USABLE state, and stay there */
if (drc->type == SPAPR_DR_CONNECTOR_TYPE_PCI) {
drc->allocation_state = SPAPR_DR_ALLOCATION_STATE_USABLE;
}
return drc;
}
static void spapr_dr_connector_instance_init(Object *obj)
{
sPAPRDRConnector *drc = SPAPR_DR_CONNECTOR(obj);
object_property_add_uint32_ptr(obj, "isolation-state",
&drc->isolation_state, NULL);
object_property_add_uint32_ptr(obj, "indicator-state",
&drc->indicator_state, NULL);
object_property_add_uint32_ptr(obj, "allocation-state",
&drc->allocation_state, NULL);
object_property_add_uint32_ptr(obj, "id", &drc->id, NULL);
object_property_add(obj, "index", "uint32", prop_get_index,
NULL, NULL, NULL, NULL);
object_property_add(obj, "connector_type", "uint32", prop_get_type,
NULL, NULL, NULL, NULL);
object_property_add_str(obj, "name", prop_get_name, NULL, NULL);
object_property_add(obj, "entity-sense", "uint32", prop_get_entity_sense,
NULL, NULL, NULL, NULL);
object_property_add(obj, "fdt", "struct", prop_get_fdt,
NULL, NULL, NULL, NULL);
}
static void spapr_dr_connector_class_init(ObjectClass *k, void *data)
{
DeviceClass *dk = DEVICE_CLASS(k);
sPAPRDRConnectorClass *drck = SPAPR_DR_CONNECTOR_CLASS(k);
dk->reset = reset;
dk->realize = realize;
dk->unrealize = unrealize;
drck->set_isolation_state = set_isolation_state;
drck->set_indicator_state = set_indicator_state;
drck->set_allocation_state = set_allocation_state;
drck->get_index = get_index;
drck->get_type = get_type;
drck->get_name = get_name;
drck->get_fdt = get_fdt;
drck->set_configured = set_configured;
drck->entity_sense = entity_sense;
drck->attach = attach;
drck->detach = detach;
drck->release_pending = release_pending;
spapr_drc: enable immediate detach for unsignalled devices Currently spapr doesn't support "aborting" hotplug of PCI devices by allowing device_del to immediately remove the device if we haven't signalled the presence of the device to the guest. In the past this wasn't an issue, since we always immediately signalled device attach and simply relied on full guest-aware add->remove path for device removal. However, as of 788d259, we now defer signalling for PCI functions until function 0 is attached, so now we need to deal with these "abort" operations for cases where a user hotplugs a non-0 function, then opts to remove it prior hotplugging function 0. Currently they'd have to reboot before the unplug completed. PCIe multifunction hotplug does not have this requirement however, so from a management implementation perspective it would be good to address this within the same release as 788d259. We accomplish this by simply adding a 'signalled' flag to track whether a device hotplug event has been sent to the guest. If it hasn't, we allow immediate removal under the assumption that the guest will not be using the device. Devices present at boot/reset time are also assumed to be 'signalled'. For CPU/memory/etc, signalling will still happen immediately as part of device_add, so only PCI functions should be affected. Cc: bharata@linux.vnet.ibm.com Cc: david@gibson.dropbear.id.au Cc: sbhat@linux.vnet.ibm.com Cc: qemu-ppc@nongnu.org Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com> [dwg: This fixes a regression where an incorrect hot-add of a non-zero function can no longer be backed out until function 0 is added] Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2016-04-01 01:27:37 +03:00
drck->set_signalled = set_signalled;
/*
* Reason: it crashes FIXME find and document the real reason
*/
qdev: Replace cannot_instantiate_with_device_add_yet with !user_creatable cannot_instantiate_with_device_add_yet was introduced by commit efec3dd631d94160288392721a5f9c39e50fb2bc to replace no_user. It was supposed to be a temporary measure. When it was introduced, we had 54 cannot_instantiate_with_device_add_yet=true lines in the code. Today (3 years later) this number has not shrunk: we now have 57 cannot_instantiate_with_device_add_yet=true lines. I think it is safe to say it is not a temporary measure, and we won't see the flag go away soon. Instead of a long field name that misleads people to believe it is temporary, replace it a shorter and less misleading field: user_creatable. Except for code comments, changes were generated using the following Coccinelle patch: @@ expression DC; @@ ( -DC->cannot_instantiate_with_device_add_yet = false; +DC->user_creatable = true; | -DC->cannot_instantiate_with_device_add_yet = true; +DC->user_creatable = false; ) @@ typedef ObjectClass; expression dc; identifier class, data; @@ static void device_class_init(ObjectClass *class, void *data) { ... dc->hotpluggable = true; +dc->user_creatable = true; ... } @@ @@ struct DeviceClass { ... -bool cannot_instantiate_with_device_add_yet; +bool user_creatable; ... } @@ expression DC; @@ ( -!DC->cannot_instantiate_with_device_add_yet +DC->user_creatable | -DC->cannot_instantiate_with_device_add_yet +!DC->user_creatable ) Cc: Alistair Francis <alistair.francis@xilinx.com> Cc: Laszlo Ersek <lersek@redhat.com> Cc: Marcel Apfelbaum <marcel@redhat.com> Cc: Markus Armbruster <armbru@redhat.com> Cc: Peter Maydell <peter.maydell@linaro.org> Cc: Thomas Huth <thuth@redhat.com> Acked-by: Alistair Francis <alistair.francis@xilinx.com> Reviewed-by: Thomas Huth <thuth@redhat.com> Reviewed-by: Marcel Apfelbaum <marcel@redhat.com> Acked-by: Marcel Apfelbaum <marcel@redhat.com> Signed-off-by: Eduardo Habkost <ehabkost@redhat.com> Message-Id: <20170503203604.31462-2-ehabkost@redhat.com> [ehabkost: kept "TODO remove once we're there" comment] Reviewed-by: Markus Armbruster <armbru@redhat.com> Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
2017-05-03 23:35:44 +03:00
dk->user_creatable = false;
}
static const TypeInfo spapr_dr_connector_info = {
.name = TYPE_SPAPR_DR_CONNECTOR,
.parent = TYPE_DEVICE,
.instance_size = sizeof(sPAPRDRConnector),
.instance_init = spapr_dr_connector_instance_init,
.class_size = sizeof(sPAPRDRConnectorClass),
.class_init = spapr_dr_connector_class_init,
};
static void spapr_drc_register_types(void)
{
type_register_static(&spapr_dr_connector_info);
}
type_init(spapr_drc_register_types)
/* helper functions for external users */
sPAPRDRConnector *spapr_dr_connector_by_index(uint32_t index)
{
Object *obj;
char name[256];
snprintf(name, sizeof(name), "%s/%x", DRC_CONTAINER_PATH, index);
obj = object_resolve_path(name, NULL);
return !obj ? NULL : SPAPR_DR_CONNECTOR(obj);
}
sPAPRDRConnector *spapr_dr_connector_by_id(sPAPRDRConnectorType type,
uint32_t id)
{
return spapr_dr_connector_by_index(
(get_type_shift(type) << DRC_INDEX_TYPE_SHIFT) |
(id & DRC_INDEX_ID_MASK));
}
/* generate a string the describes the DRC to encode into the
* device tree.
*
* as documented by PAPR+ v2.7, 13.5.2.6 and C.6.1
*/
static const char *spapr_drc_get_type_str(sPAPRDRConnectorType type)
{
switch (type) {
case SPAPR_DR_CONNECTOR_TYPE_CPU:
return "CPU";
case SPAPR_DR_CONNECTOR_TYPE_PHB:
return "PHB";
case SPAPR_DR_CONNECTOR_TYPE_VIO:
return "SLOT";
case SPAPR_DR_CONNECTOR_TYPE_PCI:
return "28";
case SPAPR_DR_CONNECTOR_TYPE_LMB:
return "MEM";
default:
g_assert(false);
}
return NULL;
}
/**
* spapr_drc_populate_dt
*
* @fdt: libfdt device tree
* @path: path in the DT to generate properties
* @owner: parent Object/DeviceState for which to generate DRC
* descriptions for
* @drc_type_mask: mask of sPAPRDRConnectorType values corresponding
* to the types of DRCs to generate entries for
*
* generate OF properties to describe DRC topology/indices to guests
*
* as documented in PAPR+ v2.1, 13.5.2
*/
int spapr_drc_populate_dt(void *fdt, int fdt_offset, Object *owner,
uint32_t drc_type_mask)
{
Object *root_container;
ObjectProperty *prop;
ObjectPropertyIterator iter;
uint32_t drc_count = 0;
GArray *drc_indexes, *drc_power_domains;
GString *drc_names, *drc_types;
int ret;
/* the first entry of each properties is a 32-bit integer encoding
* the number of elements in the array. we won't know this until
* we complete the iteration through all the matching DRCs, but
* reserve the space now and set the offsets accordingly so we
* can fill them in later.
*/
drc_indexes = g_array_new(false, true, sizeof(uint32_t));
drc_indexes = g_array_set_size(drc_indexes, 1);
drc_power_domains = g_array_new(false, true, sizeof(uint32_t));
drc_power_domains = g_array_set_size(drc_power_domains, 1);
drc_names = g_string_set_size(g_string_new(NULL), sizeof(uint32_t));
drc_types = g_string_set_size(g_string_new(NULL), sizeof(uint32_t));
/* aliases for all DRConnector objects will be rooted in QOM
* composition tree at DRC_CONTAINER_PATH
*/
root_container = container_get(object_get_root(), DRC_CONTAINER_PATH);
object_property_iter_init(&iter, root_container);
while ((prop = object_property_iter_next(&iter))) {
Object *obj;
sPAPRDRConnector *drc;
sPAPRDRConnectorClass *drck;
uint32_t drc_index, drc_power_domain;
if (!strstart(prop->type, "link<", NULL)) {
continue;
}
obj = object_property_get_link(root_container, prop->name, NULL);
drc = SPAPR_DR_CONNECTOR(obj);
drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
if (owner && (drc->owner != owner)) {
continue;
}
if ((drc->type & drc_type_mask) == 0) {
continue;
}
drc_count++;
/* ibm,drc-indexes */
drc_index = cpu_to_be32(drck->get_index(drc));
g_array_append_val(drc_indexes, drc_index);
/* ibm,drc-power-domains */
drc_power_domain = cpu_to_be32(-1);
g_array_append_val(drc_power_domains, drc_power_domain);
/* ibm,drc-names */
drc_names = g_string_append(drc_names, drck->get_name(drc));
drc_names = g_string_insert_len(drc_names, -1, "\0", 1);
/* ibm,drc-types */
drc_types = g_string_append(drc_types,
spapr_drc_get_type_str(drc->type));
drc_types = g_string_insert_len(drc_types, -1, "\0", 1);
}
/* now write the drc count into the space we reserved at the
* beginning of the arrays previously
*/
*(uint32_t *)drc_indexes->data = cpu_to_be32(drc_count);
*(uint32_t *)drc_power_domains->data = cpu_to_be32(drc_count);
*(uint32_t *)drc_names->str = cpu_to_be32(drc_count);
*(uint32_t *)drc_types->str = cpu_to_be32(drc_count);
ret = fdt_setprop(fdt, fdt_offset, "ibm,drc-indexes",
drc_indexes->data,
drc_indexes->len * sizeof(uint32_t));
if (ret) {
error_report("Couldn't create ibm,drc-indexes property");
goto out;
}
ret = fdt_setprop(fdt, fdt_offset, "ibm,drc-power-domains",
drc_power_domains->data,
drc_power_domains->len * sizeof(uint32_t));
if (ret) {
error_report("Couldn't finalize ibm,drc-power-domains property");
goto out;
}
ret = fdt_setprop(fdt, fdt_offset, "ibm,drc-names",
drc_names->str, drc_names->len);
if (ret) {
error_report("Couldn't finalize ibm,drc-names property");
goto out;
}
ret = fdt_setprop(fdt, fdt_offset, "ibm,drc-types",
drc_types->str, drc_types->len);
if (ret) {
error_report("Couldn't finalize ibm,drc-types property");
goto out;
}
out:
g_array_free(drc_indexes, true);
g_array_free(drc_power_domains, true);
g_string_free(drc_names, true);
g_string_free(drc_types, true);
return ret;
}