From: Alexey Kardashevskiy <aik@ozlabs.ru>
To: Alexander Graf <agraf@suse.de>, qemu-devel@nongnu.org
Cc: "Paolo Bonzini" <pbonzini@redhat.com>,
qemu-ppc@nongnu.org, "Andreas Färber" <afaerber@suse.de>,
"Juan Quintela" <quintela@redhat.com>
Subject: Re: [Qemu-devel] [PATCH 1/8] spapr-iommu: add a bus for spapr-iommu devices
Date: Fri, 11 Apr 2014 00:40:51 +1000 [thread overview]
Message-ID: <5346AD73.9040506@ozlabs.ru> (raw)
In-Reply-To: <53469153.9090309@suse.de>
On 04/10/2014 10:40 PM, Alexander Graf wrote:
>
> On 14.03.14 05:18, Alexey Kardashevskiy wrote:
>> At the moment sPAPR IOMMU table is a device which participates in
>> a migration stream. Normally QEMU uses a get_dev_path() hook from
>> the device's bus to compose the section name and @instance_id which are
>> used to match the section to the real device. This works till the user
>> changes the device order in the command line - if this happens,
>> devices get other instance_id's and migration fails.
>>
>> This adds a TCE bridge bus device per sPAPR machine and places all sPAPR
>> IOMMU devices onto it.
>>
>> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
>
> Juan, is a different command line device order supposed to work with
> migration?
We discussed this on IRC with Paolo and the conclusion is that yes, the
order should not matter.
Ideally we should implement "irq" property for every device (and INTA/B/C/D
for PHB) and run the source and destination QEMU with exact IRQ numbers (of
nail IRQ numbers to devices somehow?). But for me either is overkill.
>
> Alex
>
>> ---
>> hw/ppc/spapr.c | 3 +++
>> hw/ppc/spapr_iommu.c | 59
>> +++++++++++++++++++++++++++++++++++++++++++++++++-
>> include/hw/ppc/spapr.h | 7 ++++++
>> 3 files changed, 68 insertions(+), 1 deletion(-)
>>
>> diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
>> index 5c9a154..12adc21 100644
>> --- a/hw/ppc/spapr.c
>> +++ b/hw/ppc/spapr.c
>> @@ -1263,6 +1263,9 @@ static void ppc_spapr_init(QEMUMachineInitArgs *args)
>> /* Set up EPOW events infrastructure */
>> spapr_events_init(spapr);
>> + /* Set up TCE IOMMUs bus */
>> + spapr->tce_bus = spapr_tce_bus_init();
>> +
>> /* Set up VIO bus */
>> spapr->vio_bus = spapr_vio_bus_init();
>> diff --git a/hw/ppc/spapr_iommu.c b/hw/ppc/spapr_iommu.c
>> index d9fe946..7db0acf 100644
>> --- a/hw/ppc/spapr_iommu.c
>> +++ b/hw/ppc/spapr_iommu.c
>> @@ -157,7 +157,7 @@ sPAPRTCETable *spapr_tce_new_table(DeviceState
>> *owner, uint32_t liobn, size_t wi
>> return NULL;
>> }
>> - tcet = SPAPR_TCE_TABLE(object_new(TYPE_SPAPR_TCE_TABLE));
>> + tcet = SPAPR_TCE_TABLE(qdev_create(spapr->tce_bus,
>> TYPE_SPAPR_TCE_TABLE));
>> tcet->liobn = liobn;
>> tcet->window_size = window_size;
>> @@ -342,9 +342,66 @@ static TypeInfo spapr_tce_table_info = {
>> .instance_finalize = spapr_tce_table_finalize,
>> };
>> +static char *spapr_tce_bus_get_dev_name(DeviceState *qdev)
>> +{
>> + sPAPRTCETable *tcet = SPAPR_TCE_TABLE(qdev);
>> + char *name;
>> +
>> + name = g_strdup_printf("liobn@%x", tcet->liobn);
>> + return name;
>> +}
>> +
>> +static void spapr_tce_bus_class_init(ObjectClass *klass, void *data)
>> +{
>> + BusClass *k = BUS_CLASS(klass);
>> +
>> + k->get_dev_path = spapr_tce_bus_get_dev_name;
>> +}
>> +
>> +static const TypeInfo spapr_tce_bus_info = {
>> + .name = TYPE_SPAPR_TCE_BUS,
>> + .parent = TYPE_BUS,
>> + .class_init = spapr_tce_bus_class_init,
>> + .instance_size = sizeof(BusState),
>> +};
>> +
>> +static int spapr_tce_bridge_init(SysBusDevice *dev)
>> +{
>> + /* nothing */
>> + return 0;
>> +}
>> +
>> +static void spapr_tce_bridge_class_init(ObjectClass *klass, void *data)
>> +{
>> + SysBusDeviceClass *k = SYS_BUS_DEVICE_CLASS(klass);
>> +
>> + k->init = spapr_tce_bridge_init;
>> +}
>> +
>> +static const TypeInfo spapr_tce_bridge_info = {
>> + .name = "spapr-tce-bridge",
>> + .parent = TYPE_SYS_BUS_DEVICE,
>> + .instance_size = sizeof(SysBusDevice),
>> + .class_init = spapr_tce_bridge_class_init,
>> +};
>> +
>> static void register_types(void)
>> {
>> type_register_static(&spapr_tce_table_info);
>> + type_register_static(&spapr_tce_bridge_info);
>> + type_register_static(&spapr_tce_bus_info);
>> +}
>> +
>> +BusState *spapr_tce_bus_init(void)
>> +{
>> + DeviceState *dev;
>> +
>> + /* Create bridge device */
>> + dev = qdev_create(NULL, spapr_tce_bridge_info.name);
>> + qdev_init_nofail(dev);
>> +
>> + /* Create bus on bridge device */
>> + return qbus_create(TYPE_SPAPR_TCE_BUS, dev, "spapr-tce");
>> }
>> type_init(register_types);
>> diff --git a/include/hw/ppc/spapr.h b/include/hw/ppc/spapr.h
>> index 449fc7c..18332fd 100644
>> --- a/include/hw/ppc/spapr.h
>> +++ b/include/hw/ppc/spapr.h
>> @@ -12,6 +12,7 @@ struct sPAPRNVRAM;
>> typedef struct sPAPREnvironment {
>> struct VIOsPAPRBus *vio_bus;
>> + BusState *tce_bus;
>> QLIST_HEAD(, sPAPRPHBState) phbs;
>> hwaddr msi_win_addr;
>> MemoryRegion msiwindow;
>> @@ -405,4 +406,10 @@ int spapr_dma_dt(void *fdt, int node_off, const char
>> *propname,
>> int spapr_tcet_dma_dt(void *fdt, int node_off, const char *propname,
>> sPAPRTCETable *tcet);
>> +#define TYPE_SPAPR_TCE_BUS "spapr-tce-bus"
>> +#define SPAPR_TCE_BUS(obj) \
>> + OBJECT_CHECK(BusState, (obj), TYPE_SPAPR_TCE_BUS)
>> +
>> +BusState *spapr_tce_bus_init(void);
>> +
>> #endif /* !defined (__HW_SPAPR_H__) */
>
--
Alexey
next prev parent reply other threads:[~2014-04-10 14:41 UTC|newest]
Thread overview: 35+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-03-14 4:18 [Qemu-devel] [PATCH 0/8] spapr: fix IOMMU and XICS/IRQs migration Alexey Kardashevskiy
2014-03-14 4:18 ` [Qemu-devel] [PATCH 1/8] spapr-iommu: add a bus for spapr-iommu devices Alexey Kardashevskiy
2014-04-10 12:40 ` Alexander Graf
2014-04-10 14:40 ` Alexey Kardashevskiy [this message]
2014-04-10 14:52 ` Andreas Färber
2014-04-10 15:18 ` Alexey Kardashevskiy
2014-03-14 4:18 ` [Qemu-devel] [PATCH 2/8] xics: add flags for interrupts Alexey Kardashevskiy
2014-04-10 12:43 ` Alexander Graf
2014-03-14 4:18 ` [Qemu-devel] [PATCH 3/8] xics: add find_server Alexey Kardashevskiy
2014-03-14 4:18 ` [Qemu-devel] [PATCH 4/8] xics: add pre_load() hook to ICSStateClass Alexey Kardashevskiy
2014-03-14 4:18 ` [Qemu-devel] [PATCH 5/8] xics: disable flags reset on xics reset Alexey Kardashevskiy
2014-03-14 4:18 ` [Qemu-devel] [PATCH 6/8] spapr: move interrupt allocator to xics Alexey Kardashevskiy
2014-04-10 12:51 ` Alexander Graf
2014-04-10 13:24 ` Alexey Kardashevskiy
2014-04-10 13:26 ` Alexander Graf
2014-04-10 14:43 ` Alexey Kardashevskiy
2014-04-11 9:24 ` Alexander Graf
2014-04-11 12:38 ` Alexey Kardashevskiy
2014-04-11 13:58 ` Alexander Graf
2014-04-11 14:50 ` Alexey Kardashevskiy
2014-04-11 14:58 ` Alexander Graf
2014-04-11 15:27 ` Alexey Kardashevskiy
2014-04-11 15:38 ` Alexander Graf
2014-04-11 16:01 ` Alexey Kardashevskiy
2014-04-11 16:15 ` Alexander Graf
2014-04-11 16:30 ` Alexey Kardashevskiy
2014-03-14 4:18 ` [Qemu-devel] [PATCH 7/8] spapr: remove @next_irq Alexey Kardashevskiy
2014-03-14 7:19 ` Thomas Huth
2014-03-14 4:18 ` [Qemu-devel] [PATCH 8/8] xics: enable interrupt configuration reset on migration Alexey Kardashevskiy
2014-04-10 12:55 ` Alexander Graf
2014-03-20 1:25 ` [Qemu-devel] [PATCH 0/8] spapr: fix IOMMU and XICS/IRQs migration Andreas Färber
2014-04-04 5:53 ` Alexey Kardashevskiy
2014-05-04 13:56 ` Alexey Kardashevskiy
2014-05-04 21:52 ` Paolo Bonzini
2014-05-04 23:48 ` Alexey Kardashevskiy
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=5346AD73.9040506@ozlabs.ru \
--to=aik@ozlabs.ru \
--cc=afaerber@suse.de \
--cc=agraf@suse.de \
--cc=pbonzini@redhat.com \
--cc=qemu-devel@nongnu.org \
--cc=qemu-ppc@nongnu.org \
--cc=quintela@redhat.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).