* [dpdk-dev] [PATCH] net/virtio: fix secondary process crash with PCI devices
@ 2021-02-03 15:58 Maxime Coquelin
2021-02-03 17:03 ` David Marchand
2021-02-03 17:19 ` Maxime Coquelin
0 siblings, 2 replies; 3+ messages in thread
From: Maxime Coquelin @ 2021-02-03 15:58 UTC (permalink / raw)
To: dev, yinan.wang, chenbo.xia, amorenoz, david.marchand, weix.ling,
yux.jiang, anatoly.burakov
Cc: Maxime Coquelin
The Virtio rework series mistakenly moved the rte_pci_device
pointer to struct virtio_hw, which is shared between the two
processes. But this structure is per-process, so this change
made secondary process to try accessing primary process-only
memory, leading to a crash
This patch reverts to proper behavior, by storing the
rte_pci_device pointer into the pre-process
virtio_pci_internal struct. It provides also helper to get
the pointer from the virtio_hw struct pointer.
Bugzilla ID: 633
Fixes: c8d4b02f72ae ("net/virtio: move legacy IO to virtio PCI")
Reported-by: Anatoly Burakov <anatoly.burakov@intel.com>
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
drivers/net/virtio/virtio_pci.c | 25 +++++--------------------
drivers/net/virtio/virtio_pci.h | 12 +++++++++++-
drivers/net/virtio/virtio_pci_ethdev.c | 2 ++
3 files changed, 18 insertions(+), 21 deletions(-)
diff --git a/drivers/net/virtio/virtio_pci.c b/drivers/net/virtio/virtio_pci.c
index d27e9eb515..c14d1339c9 100644
--- a/drivers/net/virtio/virtio_pci.c
+++ b/drivers/net/virtio/virtio_pci.c
@@ -31,13 +31,6 @@
#define VIRTIO_PCI_CONFIG(dev) \
(((dev)->msix_status == VIRTIO_MSIX_ENABLED) ? 24 : 20)
-
-struct virtio_pci_internal {
- struct rte_pci_ioport io;
-};
-
-#define VTPCI_IO(hw) (&virtio_pci_internal[(hw)->port_id].io)
-
struct virtio_pci_internal virtio_pci_internal[RTE_MAX_ETHPORTS];
static inline int
@@ -313,16 +306,14 @@ legacy_intr_detect(struct virtio_hw *hw)
{
struct virtio_pci_dev *dev = virtio_pci_get_dev(hw);
- dev->msix_status = vtpci_msix_detect(dev->pci_dev);
+ dev->msix_status = vtpci_msix_detect(VTPCI_DEV(hw));
hw->intr_lsc = !!dev->msix_status;
}
static int
legacy_dev_close(struct virtio_hw *hw)
{
- struct virtio_pci_dev *dev = virtio_pci_get_dev(hw);
-
- rte_pci_unmap_device(dev->pci_dev);
+ rte_pci_unmap_device(VTPCI_DEV(hw));
rte_pci_ioport_unmap(VTPCI_IO(hw));
return 0;
@@ -574,16 +565,14 @@ modern_intr_detect(struct virtio_hw *hw)
{
struct virtio_pci_dev *dev = virtio_pci_get_dev(hw);
- dev->msix_status = vtpci_msix_detect(dev->pci_dev);
+ dev->msix_status = vtpci_msix_detect(VTPCI_DEV(hw));
hw->intr_lsc = !!dev->msix_status;
}
static int
modern_dev_close(struct virtio_hw *hw)
{
- struct virtio_pci_dev *dev = virtio_pci_get_dev(hw);
-
- rte_pci_unmap_device(dev->pci_dev);
+ rte_pci_unmap_device(VTPCI_DEV(hw));
return 0;
}
@@ -772,8 +761,6 @@ vtpci_init(struct rte_pci_device *pci_dev, struct virtio_pci_dev *dev)
RTE_BUILD_BUG_ON(offsetof(struct virtio_pci_dev, hw) != 0);
- dev->pci_dev = pci_dev;
-
/*
* Try if we can succeed reading virtio pci caps, which exists
* only on modern pci device. If failed, we fallback to legacy
@@ -816,7 +803,5 @@ void vtpci_legacy_ioport_unmap(struct virtio_hw *hw)
int vtpci_legacy_ioport_map(struct virtio_hw *hw)
{
- struct virtio_pci_dev *dev = virtio_pci_get_dev(hw);
-
- return rte_pci_ioport_map(dev->pci_dev, 0, VTPCI_IO(hw));
+ return rte_pci_ioport_map(VTPCI_DEV(hw), 0, VTPCI_IO(hw));
}
diff --git a/drivers/net/virtio/virtio_pci.h b/drivers/net/virtio/virtio_pci.h
index 57596e471f..11e25a0142 100644
--- a/drivers/net/virtio/virtio_pci.h
+++ b/drivers/net/virtio/virtio_pci.h
@@ -104,7 +104,6 @@ enum virtio_msix_status {
struct virtio_pci_dev {
struct virtio_hw hw;
- struct rte_pci_device *pci_dev;
struct virtio_pci_common_cfg *common_cfg;
struct virtio_net_config *dev_cfg;
enum virtio_msix_status msix_status;
@@ -116,6 +115,17 @@ struct virtio_pci_dev {
#define virtio_pci_get_dev(hwp) container_of(hwp, struct virtio_pci_dev, hw)
+struct virtio_pci_internal {
+ struct rte_pci_ioport io;
+ struct rte_pci_device *dev;
+};
+
+extern struct virtio_pci_internal virtio_pci_internal[RTE_MAX_ETHPORTS];
+
+#define VTPCI_IO(hw) (&virtio_pci_internal[(hw)->port_id].io)
+#define VTPCI_DEV(hw) (virtio_pci_internal[(hw)->port_id].dev)
+
+
/*
* How many bits to shift physical queue address written to QUEUE_PFN.
* 12 is historical, and due to x86 page size.
diff --git a/drivers/net/virtio/virtio_pci_ethdev.c b/drivers/net/virtio/virtio_pci_ethdev.c
index 3bb5c6268b..4083853c48 100644
--- a/drivers/net/virtio/virtio_pci_ethdev.c
+++ b/drivers/net/virtio/virtio_pci_ethdev.c
@@ -78,12 +78,14 @@ eth_virtio_pci_init(struct rte_eth_dev *eth_dev)
if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
hw->port_id = eth_dev->data->port_id;
+ VTPCI_DEV(hw) = pci_dev;
ret = vtpci_init(RTE_ETH_DEV_TO_PCI(eth_dev), dev);
if (ret) {
PMD_INIT_LOG(ERR, "Failed to init PCI device\n");
return -1;
}
} else {
+ VTPCI_DEV(hw) = pci_dev;
if (dev->modern)
VIRTIO_OPS(hw) = &modern_ops;
else
--
2.29.2
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [dpdk-dev] [PATCH] net/virtio: fix secondary process crash with PCI devices
2021-02-03 15:58 [dpdk-dev] [PATCH] net/virtio: fix secondary process crash with PCI devices Maxime Coquelin
@ 2021-02-03 17:03 ` David Marchand
2021-02-03 17:19 ` Maxime Coquelin
1 sibling, 0 replies; 3+ messages in thread
From: David Marchand @ 2021-02-03 17:03 UTC (permalink / raw)
To: Maxime Coquelin
Cc: dev, Wang, Yinan, Xia, Chenbo, Adrian Moreno Zapata, Wei Ling,
Yu Jiang, Burakov, Anatoly
On Wed, Feb 3, 2021 at 4:58 PM Maxime Coquelin
<maxime.coquelin@redhat.com> wrote:
>
> The Virtio rework series mistakenly moved the rte_pci_device
> pointer to struct virtio_hw, which is shared between the two
> processes. But this structure is per-process, so this change
> made secondary process to try accessing primary process-only
> memory, leading to a crash
crash.*
Ah the joy of multiprocess support :-).
>
> This patch reverts to proper behavior, by storing the
> rte_pci_device pointer into the pre-process
per*
> virtio_pci_internal struct. It provides also helper to get
also provides*
> the pointer from the virtio_hw struct pointer.
>
> Bugzilla ID: 633
> Fixes: c8d4b02f72ae ("net/virtio: move legacy IO to virtio PCI")
>
> Reported-by: Anatoly Burakov <anatoly.burakov@intel.com>
> Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
The patch lgtm.
Reviewed-by: David Marchand <david.marchand@redhat.com>
Thanks.
--
David Marchand
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [dpdk-dev] [PATCH] net/virtio: fix secondary process crash with PCI devices
2021-02-03 15:58 [dpdk-dev] [PATCH] net/virtio: fix secondary process crash with PCI devices Maxime Coquelin
2021-02-03 17:03 ` David Marchand
@ 2021-02-03 17:19 ` Maxime Coquelin
1 sibling, 0 replies; 3+ messages in thread
From: Maxime Coquelin @ 2021-02-03 17:19 UTC (permalink / raw)
To: dev, yinan.wang, chenbo.xia, amorenoz, david.marchand, weix.ling,
yux.jiang, anatoly.burakov
On 2/3/21 4:58 PM, Maxime Coquelin wrote:
> The Virtio rework series mistakenly moved the rte_pci_device
> pointer to struct virtio_hw, which is shared between the two
> processes. But this structure is per-process, so this change
> made secondary process to try accessing primary process-only
> memory, leading to a crash
>
> This patch reverts to proper behavior, by storing the
> rte_pci_device pointer into the pre-process
> virtio_pci_internal struct. It provides also helper to get
> the pointer from the virtio_hw struct pointer.
>
> Bugzilla ID: 633
> Fixes: c8d4b02f72ae ("net/virtio: move legacy IO to virtio PCI")
>
> Reported-by: Anatoly Burakov <anatoly.burakov@intel.com>
> Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
> ---
> drivers/net/virtio/virtio_pci.c | 25 +++++--------------------
> drivers/net/virtio/virtio_pci.h | 12 +++++++++++-
> drivers/net/virtio/virtio_pci_ethdev.c | 2 ++
> 3 files changed, 18 insertions(+), 21 deletions(-)
>
Applied to dpdk-next-virtio/main with typo fixed.
Thanks,
Maxime
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2021-02-03 17:19 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-02-03 15:58 [dpdk-dev] [PATCH] net/virtio: fix secondary process crash with PCI devices Maxime Coquelin
2021-02-03 17:03 ` David Marchand
2021-02-03 17:19 ` Maxime Coquelin
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).