[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Qemu-devel] Re: [PATCH v7 3/6] pcie/aer: glue aer error injection into
From: |
Michael S. Tsirkin |
Subject: |
[Qemu-devel] Re: [PATCH v7 3/6] pcie/aer: glue aer error injection into qemu monitor |
Date: |
Tue, 2 Nov 2010 14:10:28 +0200 |
User-agent: |
Mutt/1.5.21 (2010-09-15) |
On Tue, Nov 02, 2010 at 06:32:49PM +0900, Isaku Yamahata wrote:
> introduce pcie_aer_inject_error command.
>
> Signed-off-by: Isaku Yamahata <address@hidden>
> ---
> Changes v6 -> v7:
> - check return value.
>
> Changes v3 -> v4:
> - s/PCIE_AER/PCIEAER/g for structure names.
> - compilation adjustment.
>
> Changes v2 -> v3:
> - compilation adjustment.
> ---
> hmp-commands.hx | 23 ++++++++++++++
> hw/pcie_aer.c | 87
> +++++++++++++++++++++++++++++++++++++++++++++++++++++++
> sysemu.h | 5 +++
> 3 files changed, 115 insertions(+), 0 deletions(-)
>
> diff --git a/hmp-commands.hx b/hmp-commands.hx
> index 81999aa..02eae7d 100644
> --- a/hmp-commands.hx
> +++ b/hmp-commands.hx
> @@ -835,6 +835,29 @@ Hot remove PCI device.
> ETEXI
>
> {
> + .name = "pcie_aer_inject_error",
> + .args_type = "advisory_non_fatal:-a,is_correctable:-c,"
is_correctable->correctable?
> + "pci_addr:s,error_status:i,"
> + "tlph0:i?,tlph1:i?,tlph2:i?,tlph3:i?,"
> + "hpfx0:i?,hpfx1:i?,hpfx2:i?,hpfx3:i?",
Do the field names have to be so cryptic?
> + .params = "[-a] [-c] [[<domain>:]<bus>:]<slot>.<func> "
> + "<error status:32bit> "
Only 1 bit is valid though, right?
So I would say let's pass symbolic name, not raw bit.
> + "[<tlp header:(32bit x 4)>] "
> + "[<tlp header prefix:(32bit x 4)>]",
Maybe -advisory -correctable?
> + .help = "inject pcie aer error "
> + "(use -a for advisory non fatal error) "
> + "(use -c for correctrable error)",
> + .user_print = pcie_aer_inject_error_print,
> + .mhandler.cmd_new = do_pcie_aer_inejct_error,
> + },
> +
> +STEXI
> address@hidden pcie_aer_inject_error
> address@hidden pcie_aer_inject_error
> +Inject PCIe AER error
> +ETEXI
> +
> + {
> .name = "host_net_add",
> .args_type = "device:s,opts:s?",
> .params = "tap|user|socket|vde|dump [options]",
> diff --git a/hw/pcie_aer.c b/hw/pcie_aer.c
> index 84c3457..30e39fd 100644
> --- a/hw/pcie_aer.c
> +++ b/hw/pcie_aer.c
> @@ -19,6 +19,8 @@
> */
>
> #include "sysemu.h"
> +#include "qemu-objects.h"
> +#include "monitor.h"
> #include "pci_bridge.h"
> #include "pcie.h"
> #include "msix.h"
> @@ -823,3 +825,88 @@ const VMStateDescription vmstate_pcie_aer_log = {
> }
> };
>
> +void pcie_aer_inject_error_print(Monitor *mon, const QObject *data)
> +{
> + QDict *qdict;
> + int devfn;
> + assert(qobject_type(data) == QTYPE_QDICT);
> + qdict = qobject_to_qdict(data);
> +
> + devfn = (int)qdict_get_int(qdict, "devfn");
> + monitor_printf(mon, "OK domain: %x, bus: %x devfn: %x.%x\n",
> + (int) qdict_get_int(qdict, "domain"),
> + (int) qdict_get_int(qdict, "bus"),
> + PCI_SLOT(devfn), PCI_FUNC(devfn));
> +}
> +
> +int do_pcie_aer_inejct_error(Monitor *mon,
> + const QDict *qdict, QObject **ret_data)
> +{
> + const char *pci_addr = qdict_get_str(qdict, "pci_addr");
> + int dom;
> + int bus;
> + unsigned int slot;
> + unsigned int func;
> + PCIDevice *dev;
> + PCIEAERErr err;
> + int ret;
> +
> + /* Ideally qdev device path should be used.
> + * However at the moment there is no reliable way to determine
> + * wheher a given qdev is pci device or not.
> + * so pci_addr is used.
> + */
This does not look like such a big deal, and tying interface
to such implementation detail seems wrong to me.
> + if (pci_parse_devaddr(pci_addr, &dom, &bus, &slot, &func)) {
> + monitor_printf(mon, "invalid pci address %s\n", pci_addr);
> + return -1;
> + }
> + dev = pci_find_device(pci_find_root_bus(dom), bus, slot, func);
> + if (!dev) {
> + monitor_printf(mon, "device is not found. 0x%x:0x%x.0x%x\n",
> + bus, slot, func);
> + return -1;
> + }
> + if (!pci_is_express(dev)) {
> + monitor_printf(mon, "the device doesn't support pci express. "
> + "0x%x:0x%x.0x%x\n",
> + bus, slot, func);
> + return -1;
> + }
> +
> + err.status = qdict_get_int(qdict, "error_status");
> + err.source_id = (pci_bus_num(dev->bus) << 8) | dev->devfn;
> +
> + err.flags = 0;
> + if (qdict_get_int(qdict, "is_correctable")) {
> + err.flags |= PCIE_AER_ERR_IS_CORRECTABLE;
> + }
> + if (qdict_get_int(qdict, "advisory_non_fatal")) {
> + err.flags |= PCIE_AER_ERR_MAYBE_ADVISORY;
> + }
> + if (qdict_haskey(qdict, "tlph0")) {
> + err.flags |= PCIE_AER_ERR_HEADER_VALID;
> + }
> + if (qdict_haskey(qdict, "hpfx0")) {
> + err.flags |= PCIE_AER_ERR_TLP_PRESENT;
> + }
So tlph0 is for header and hpfx0 is for TLP_PRESENT?
Not the other way around?
> +
> + err.header[0] = qdict_get_try_int(qdict, "tlph0", 0);
> + err.header[1] = qdict_get_try_int(qdict, "tlph1", 0);
> + err.header[2] = qdict_get_try_int(qdict, "tlph2", 0);
> + err.header[3] = qdict_get_try_int(qdict, "tlph3", 0);
> +
> + err.prefix[0] = qdict_get_try_int(qdict, "hpfx0", 0);
> + err.prefix[1] = qdict_get_try_int(qdict, "hpfx1", 0);
> + err.prefix[2] = qdict_get_try_int(qdict, "hpfx2", 0);
> + err.prefix[3] = qdict_get_try_int(qdict, "hpfx3", 0);
> +
> + ret = pcie_aer_inject_error(dev, &err);
> + *ret_data = qobject_from_jsonf("{ 'domain': %d, 'bus': %d, 'devfn': %d "
> + "'ret': %d}",
> + pci_find_domain(dev->bus),
> + pci_bus_num(dev->bus), dev->devfn,
> + ret);
> + assert(*ret_data);
> +
> + return 0;
> +}
> diff --git a/sysemu.h b/sysemu.h
> index b81a70e..99c7909 100644
> --- a/sysemu.h
> +++ b/sysemu.h
> @@ -151,6 +151,11 @@ void pci_device_hot_add(Monitor *mon, const QDict
> *qdict);
> void drive_hot_add(Monitor *mon, const QDict *qdict);
> void do_pci_device_hot_remove(Monitor *mon, const QDict *qdict);
>
> +/* pcie aer error injection */
> +void pcie_aer_inject_error_print(Monitor *mon, const QObject *data);
> +int do_pcie_aer_inejct_error(Monitor *mon,
> + const QDict *qdict, QObject **ret_data);
> +
> /* serial ports */
>
> #define MAX_SERIAL_PORTS 4
> --
> 1.7.1.1
[Qemu-devel] [PATCH v7 1/6] pcie_regs.h: more constants, Isaku Yamahata, 2010/11/02
[Qemu-devel] [PATCH v7 3/6] pcie/aer: glue aer error injection into qemu monitor, Isaku Yamahata, 2010/11/02
- [Qemu-devel] Re: [PATCH v7 3/6] pcie/aer: glue aer error injection into qemu monitor,
Michael S. Tsirkin <=
[Qemu-devel] [PATCH v7 5/6] x3130/upstream: support aer, Isaku Yamahata, 2010/11/02
[Qemu-devel] [PATCH v7 6/6] x3130/downstream: support aer., Isaku Yamahata, 2010/11/02
[Qemu-devel] Re: [PATCH v7 0/6] pcie port switch emulators, Michael S. Tsirkin, 2010/11/02