@@ -495,7 +495,8 @@ struct hv_pcibus_device {
spinlock_t device_list_lock; /* Protect lists below */
void __iomem *cfg_addr;
- spinlock_t idr_lock; /* Serialize accesses to the IDR */
+ /* Serialize accesses to the IDR; see also hv_pci_onchannelcallback(). */
+ spinlock_t idr_lock;
struct idr idr; /* Map guest memory addresses */
struct list_head children;
@@ -2797,16 +2798,24 @@ static void hv_pci_onchannelcallback(void *context)
}
spin_lock_irqsave(&hbus->idr_lock, flags);
comp_packet = (struct pci_packet *)idr_find(&hbus->idr, req_id);
- spin_unlock_irqrestore(&hbus->idr_lock, flags);
if (!comp_packet) {
+ spin_unlock_irqrestore(&hbus->idr_lock, flags);
dev_warn_ratelimited(&hbus->hdev->device,
"Request ID not found\n");
break;
}
response = (struct pci_response *)buffer;
+ /*
+ * Call ->completion_func() within the critical section to make
+ * sure that the packet pointer is still valid during the call:
+ * here 'valid' means that there's a task still waiting for the
+ * completion, and that the packet data is still on the waiting
+ * task's stack/has not already been freed by the waiting task.
+ */
comp_packet->completion_func(comp_packet->compl_ctxt,
response,
bytes_recvd);
+ spin_unlock_irqrestore(&hbus->idr_lock, flags);
break;
case VM_PKT_DATA_INBAND:
Dexuan wrote: "[...] when we disable AccelNet, the host PCI VSP driver sends a PCI_EJECT message first, and the channel callback may set hpdev->state to hv_pcichild_ejecting on a different CPU. This can cause hv_compose_msi_msg() to exit from the loop and 'return', and the on-stack variable 'ctxt' is invalid. Now, if the response message from the host arrives, the channel callback will try to access the invalid 'ctxt' variable, and this may cause a crash." Schematically: Hyper-V sends PCI_EJECT msg hv_pci_onchannelcallback() state = hv_pcichild_ejecting hv_compose_msi_msg() alloc and init comp_pkt state == hv_pcichild_ejecting Hyper-V sends VM_PKT_COMP msg hv_pci_onchannelcallback() retrieve address of comp_pkt 'free' comp_pkt and return comp_pkt->completion_func() Dexuan also showed how the crash can be triggered after introducing suitable delays in the driver code, thus validating the 'assumption' that the host can still normally respond to the guest's compose_msi request after the host has started to eject the PCI device. Fix the synchronization by leveraging the IDR lock. Retrieve the address of the completion packet *and call the completion function within a same critical section: if an address (request ID) is found in the channel callback, the critical section precedes the removal of the address in hv_compose_msi_msg(). Fixes: de0aa7b2f97d3 ("PCI: hv: Fix 2 hang issues in hv_compose_msi_msg()") Reported-by: Wei Hu <weh@microsoft.com> Reported-by: Dexuan Cui <decui@microsoft.com> Suggested-by: Michael Kelley <mikelley@microsoft.com> Signed-off-by: Andrea Parri (Microsoft) <parri.andrea@gmail.com> --- drivers/pci/controller/pci-hyperv.c | 13 +++++++++++-- 1 file changed, 11 insertions(+), 2 deletions(-)