qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH] qtest: add tulip test case


From: Li Qiang
Subject: Re: [PATCH] qtest: add tulip test case
Date: Mon, 30 Mar 2020 22:56:27 +0800



Jason Wang <address@hidden> 于2020年3月30日周一 下午3:25写道:

On 2020/3/28 上午12:11, Li Qiang wrote:
> The tulip networking card emulation has an OOB issue in
> 'tulip_copy_tx_buffers' when the guest provide malformed descriptor.
> This test will trigger a ASAN heap overflow crash.


Hi Qiang:

Thanks for the qtest patch.

Few nitpicks, see above.


>
> Signed-off-by: Li Qiang <address@hidden>
> ---
>   tests/qtest/Makefile.include |  1 +
>   tests/qtest/tulip-test.c     | 91 ++++++++++++++++++++++++++++++++++++
>   2 files changed, 92 insertions(+)
>   create mode 100644 tests/qtest/tulip-test.c
>
> diff --git a/tests/qtest/Makefile.include b/tests/qtest/Makefile.include
> index 10a28de8a3..9e5a51d033 100644
> --- a/tests/qtest/Makefile.include
> +++ b/tests/qtest/Makefile.include
> @@ -217,6 +217,7 @@ qos-test-obj-y += tests/qtest/es1370-test.o
>   qos-test-obj-y += tests/qtest/ipoctal232-test.o
>   qos-test-obj-y += tests/qtest/megasas-test.o
>   qos-test-obj-y += tests/qtest/ne2000-test.o
> +qos-test-obj-y += tests/qtest/tulip-test.o
>   qos-test-obj-y += tests/qtest/nvme-test.o
>   qos-test-obj-y += tests/qtest/pca9552-test.o
>   qos-test-obj-y += tests/qtest/pci-test.o
> diff --git a/tests/qtest/tulip-test.c b/tests/qtest/tulip-test.c
> new file mode 100644
> index 0000000000..d91ddfd765
> --- /dev/null
> +++ b/tests/qtest/tulip-test.c
> @@ -0,0 +1,91 @@
> +/*
> + * QTest testcase for DEC/Intel Tulip 21143
> + *
> + * Copyright (c) 2020 Li Qiang <address@hidden>
> + *
> + * 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"
> +#include "libqtest.h"
> +#include "qemu/module.h"
> +#include "libqos/qgraph.h"
> +#include "libqos/pci.h"
> +#include "qemu/bitops.h"
> +#include "hw/net/tulip.h"
> +
> +typedef struct QTulip_pci QTulip_pci;
> +
> +struct QTulip_pci {
> +    QOSGraphObject obj;
> +    QPCIDevice dev;
> +};
> +
> +static void *tulip_pci_get_driver(void *obj, const char *interface)
> +{
> +    QTulip_pci *tulip_pci = obj;
> +
> +    if (!g_strcmp0(interface, "pci-device")) {
> +        return &tulip_pci->dev;
> +    }
> +
> +    fprintf(stderr, "%s not present in tulip_pci\n", interface);
> +    g_assert_not_reached();
> +}
> +
> +static void *tulip_pci_create(void *pci_bus, QGuestAllocator *alloc, void *addr)
> +{
> +    QTulip_pci *tulip_pci = g_new0(QTulip_pci, 1);
> +    QPCIBus *bus = pci_bus;
> +
> +    qpci_device_init(&tulip_pci->dev, bus, addr);
> +    tulip_pci->obj.get_driver = tulip_pci_get_driver;
> +
> +    return &tulip_pci->obj;
> +}
> +
> +static void tulip_large_tx(void *obj, void *data, QGuestAllocator *alloc)
> +{
> +    QTulip_pci *tulip_pci = obj;
> +    QPCIDevice *dev = &tulip_pci->dev;
> +    QPCIBar bar;
> +    struct tulip_descriptor context;
> +    char guest_data[4096];
> +    uint64_t context_pa;
> +    uint64_t guest_pa;
> +
> +    qpci_device_enable(dev);
> +    bar = qpci_iomap(dev, 0, NULL);
> +    context_pa = guest_alloc(alloc, sizeof(context));
> +    guest_pa = guest_alloc(alloc, 4096);
> +    memset(guest_data, 'A', sizeof(guest_data));


It would be better to have a comment on how the descriptor is structured
to trigger the OOB.


> +    context.status = TDES0_OWN;
> +    context.control = TDES1_BUF2_SIZE_MASK << TDES1_BUF2_SIZE_SHIFT |
> +                      TDES1_BUF1_SIZE_MASK << TDES1_BUF1_SIZE_SHIFT;
> +    context.buf_addr2 = context_pa + sizeof(struct tulip_descriptor);
> +    context.buf_addr1 = guest_pa;
> +
> +    qtest_memwrite(dev->bus->qts, context_pa, &context, sizeof(context));
> +    qtest_memwrite(dev->bus->qts, guest_pa, guest_data, sizeof(guest_data));
> +    qpci_io_writel(dev, bar, 0x20, context_pa);
> +    qpci_io_writel(dev, bar, 0x30, 1 << 13);


Any chance to use macro instead of magic numbers?



Hi Jason, 
I have posted a new revision and only replace the '1<<13'. As the register has not been defined in a header file. Also I read
the kernel the register is only used in a '.c' file. So I think the '0x20' and '0x30' can use magic number. And of
course I write them as CSR4 and CSR6 in the commit message. 

Thanks,
Li Qiang

 

> +    guest_free(alloc, context_pa);
> +    guest_free(alloc, guest_pa);
> +}
> +
> +static void tulip_register_nodes(void)
> +{
> +    QOSGraphEdgeOptions opts = {
> +        .extra_device_opts = "addr=04.0",
> +    };
> +    add_qpci_address(&opts, &(QPCIAddress) { .devfn = QPCI_DEVFN(4, 0) });
> +
> +    qos_node_create_driver("tulip", tulip_pci_create);
> +    qos_node_consumes("tulip", "pci-bus", &opts);
> +    qos_node_produces("tulip", "pci-device");
> +
> +    qos_add_test("tulip_large_tx", "tulip", tulip_large_tx, NULL);
> +}
> +
> +libqos_init(tulip_register_nodes);


reply via email to

[Prev in Thread] Current Thread [Next in Thread]