similar to: [PATCH] virtio: make del_vqs idempotent

From: Longpeng <longpeng2 at huawei.com> [ Upstream commit 6a8aae68c87349dbbcd46eac380bc43cdb98a13b ] If the msix_affinity_masks is alloced failed, then we'll try to free some resources in vp_free_vectors() that may access it directly. We met the following stack in our production: [ 29.296767] BUG: unable to handle kernel NULL pointer dereference at (null) [ 29.311151] IP:

From: Longpeng <longpeng2 at huawei.com> [ Upstream commit 6a8aae68c87349dbbcd46eac380bc43cdb98a13b ] If the msix_affinity_masks is alloced failed, then we'll try to free some resources in vp_free_vectors() that may access it directly. We met the following stack in our production: [ 29.296767] BUG: unable to handle kernel NULL pointer dereference at (null) [ 29.311151] IP:

From: Longpeng <longpeng2 at huawei.com> [ Upstream commit 6a8aae68c87349dbbcd46eac380bc43cdb98a13b ] If the msix_affinity_masks is alloced failed, then we'll try to free some resources in vp_free_vectors() that may access it directly. We met the following stack in our production: [ 29.296767] BUG: unable to handle kernel NULL pointer dereference at (null) [ 29.311151] IP:

This refactors find_vqs, making it more readable and robust, and fixing two regressions from 2.6.30: - double free_irq causing BUG_ON on device removal - probe failure when vq can't be assigned to msi-x vector (reported on old host kernels) An older version of this patch was tested by Amit Shah. Reported-by: Amit Shah <amit.shah at redhat.com> Signed-off-by: Michael S. Tsirkin

This refactors find_vqs, making it more readable and robust, and fixing two regressions from 2.6.30: - double free_irq causing BUG_ON on device removal - probe failure when vq can't be assigned to msi-x vector (reported on old host kernels) An older version of this patch was tested by Amit Shah. Reported-by: Amit Shah <amit.shah at redhat.com> Signed-off-by: Michael S. Tsirkin

We don't need to go from vq to vq info on data path, so using direct vq->priv pointer for that seems like a waste. Let's build an array of vq infos, then we can use vq->index for that lookup. Signed-off-by: Michael S. Tsirkin <mst at redhat.com> --- drivers/virtio/virtio_pci.c | 56 +++++++++++++++++++++++++-------------------- 1 file changed, 31 insertions(+), 25

We don't need to go from vq to vq info on data path, so using direct vq->priv pointer for that seems like a waste. Let's build an array of vq infos, then we can use vq->index for that lookup. Signed-off-by: Michael S. Tsirkin <mst at redhat.com> --- drivers/virtio/virtio_pci.c | 56 +++++++++++++++++++++++++-------------------- 1 file changed, 31 insertions(+), 25

This refactors find_vqs, making it more readable and robust, and fixing two regressions from 2.6.30: - double free_irq causing BUG_ON on device removal - probe failure when vq can't be assigned to msi-x vector (reported on old host kernels) An older version of this patch was tested by Amit Shah. Reported-by: Amit Shah <amit.shah at redhat.com> Signed-off-by: Michael S. Tsirkin

This refactors find_vqs, making it more readable and robust, and fixing two regressions from 2.6.30: - double free_irq causing BUG_ON on device removal - probe failure when vq can't be assigned to msi-x vector (reported on old host kernels) An older version of this patch was tested by Amit Shah. Reported-by: Amit Shah <amit.shah at redhat.com> Signed-off-by: Michael S. Tsirkin

1) Rename vp_request_vectors to vp_request_msix_vectors, and take non-MSI-X case out to caller. 2) Comment weird pci_enable_msix API 3) Rename vp_find_vq to setup_vq. 4) Fix spaces to tabs 5) Make nvectors calc internal to vp_try_to_find_vqs() 6) Rename vector to msix_vector for more clarity. Signed-off-by: Rusty Russell <rusty at rustcorp.com.au> Cc: "Michael S. Tsirkin"

1) Rename vp_request_vectors to vp_request_msix_vectors, and take non-MSI-X case out to caller. 2) Comment weird pci_enable_msix API 3) Rename vp_find_vq to setup_vq. 4) Fix spaces to tabs 5) Make nvectors calc internal to vp_try_to_find_vqs() 6) Rename vector to msix_vector for more clarity. Signed-off-by: Rusty Russell <rusty at rustcorp.com.au> Cc: "Michael S. Tsirkin"

We don't need to go from vq to vq info on data path, so using direct vq->priv pointer for that seems like a waste. Let's build an array of vq infos, then we can use vq->index for that lookup. Signed-off-by: Michael S. Tsirkin <mst at redhat.com> --- drivers/virtio/virtio_pci.c | 38 +++++++++++++++++++++++--------------- 1 file changed, 23 insertions(+), 15 deletions(-)

We don't need to go from vq to vq info on data path, so using direct vq->priv pointer for that seems like a waste. Let's build an array of vq infos, then we can use vq->index for that lookup. Signed-off-by: Michael S. Tsirkin <mst at redhat.com> --- drivers/virtio/virtio_pci.c | 38 +++++++++++++++++++++++--------------- 1 file changed, 23 insertions(+), 15 deletions(-)

On (Tue) Jul 21 2009 [19:54:00], Michael S. Tsirkin wrote: > On Fri, Jul 17, 2009 at 06:34:40PM +0530, Amit Shah wrote: > > Hello, > > > > Using recent qemu-kvm userspace with a slightly older kernel module I > > get this when using the virtio-net device: I was getting this with a very recent kernel in the guest with an older host kernel. 2.6.31-rc3 in the guest and

On (Tue) Jul 21 2009 [19:54:00], Michael S. Tsirkin wrote: > On Fri, Jul 17, 2009 at 06:34:40PM +0530, Amit Shah wrote: > > Hello, > > > > Using recent qemu-kvm userspace with a slightly older kernel module I > > get this when using the virtio-net device: I was getting this with a very recent kernel in the guest with an older host kernel. 2.6.31-rc3 in the guest and

virtio currently fails to find any vqs if the device supports msi-x but fails to assign it to a queue. Turns out, this actually happens for old host kernels which can't allocate a gsi for the vector. As a result guest does not use virtio net. Fix this by retrying with less vectors and finally disabling msi on such failures and falling back on regular interrupts. Reported-by: Amit Shah

virtio currently fails to find any vqs if the device supports msi-x but fails to assign it to a queue. Turns out, this actually happens for old host kernels which can't allocate a gsi for the vector. As a result guest does not use virtio net. Fix this by retrying with less vectors and finally disabling msi on such failures and falling back on regular interrupts. Reported-by: Amit Shah

Device VQs were getting freed twice: once in every devices removal functions, and then again in virtio_pci_legacy_remove(). Signed-off-by: Sasha Levin <sasha.levin at oracle.com> --- drivers/virtio/virtio_pci_legacy.c | 1 - 1 file changed, 1 deletion(-) diff --git a/drivers/virtio/virtio_pci_legacy.c b/drivers/virtio/virtio_pci_legacy.c index 6c76f0f..913ca23 100644 ---

Device VQs were getting freed twice: once in every devices removal functions, and then again in virtio_pci_legacy_remove(). Signed-off-by: Sasha Levin <sasha.levin at oracle.com> --- drivers/virtio/virtio_pci_legacy.c | 1 - 1 file changed, 1 deletion(-) diff --git a/drivers/virtio/virtio_pci_legacy.c b/drivers/virtio/virtio_pci_legacy.c index 6c76f0f..913ca23 100644 ---

This lets IRQ layer handle dispatching IRQs to separate handlers for the case where we don't have per-VQ MSI-X vectors, and allows us to greatly simplify the code based on the assumption that we always have interrupt vector 0 (legacy INTx or config interrupt for MSI-X) available, and any other interrupt is request/freed throught the VQ, even if the actual interrupt line might be shared in some