KVM: Fix kvm_irqfd_init initialization

In commit a0f155e96 'KVM: Initialize irqfd from kvm_init()', when
kvm_init() is called the second time (e.g kvm-amd.ko and kvm-intel.ko),
kvm_arch_init() will fail with -EEXIST, then kvm_irqfd_exit() will be
called on the error handling path. This way, the kvm_irqfd system will
not be ready.

This patch fix the following:

BUG: unable to handle kernel NULL pointer dereference at           (null)
IP: [<ffffffff81c0721e>] _raw_spin_lock+0xe/0x30
PGD 0
Oops: 0002 [#1] SMP
Modules linked in: vhost_net
CPU 6
Pid: 4257, comm: qemu-system-x86 Not tainted 3.9.0-rc3+ #757 Dell Inc. OptiPlex 790/0V5HMK
RIP: 0010:[<ffffffff81c0721e>]  [<ffffffff81c0721e>] _raw_spin_lock+0xe/0x30
RSP: 0018:ffff880221721cc8  EFLAGS: 00010046
RAX: 0000000000000100 RBX: ffff88022dcc003f RCX: ffff880221734950
RDX: ffff8802208f6ca8 RSI: 000000007fffffff RDI: 0000000000000000
RBP: ffff880221721cc8 R08: 0000000000000002 R09: 0000000000000002
R10: 00007f7fd01087e0 R11: 0000000000000246 R12: ffff8802208f6ca8
R13: 0000000000000080 R14: ffff880223e2a900 R15: 0000000000000000
FS:  00007f7fd38488e0(0000) GS:ffff88022dcc0000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000000000000000 CR3: 000000022309f000 CR4: 00000000000427e0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
Process qemu-system-x86 (pid: 4257, threadinfo ffff880221720000, task ffff880222bd5640)
Stack:
 ffff880221721d08 ffffffff810ac5c5 ffff88022431dc00 0000000000000086
 0000000000000080 ffff880223e2a900 ffff8802208f6ca8 0000000000000000
 ffff880221721d48 ffffffff810ac8fe 0000000000000000 ffff880221734000
Call Trace:
 [<ffffffff810ac5c5>] __queue_work+0x45/0x2d0
 [<ffffffff810ac8fe>] queue_work_on+0x8e/0xa0
 [<ffffffff810ac949>] queue_work+0x19/0x20
 [<ffffffff81009b6b>] irqfd_deactivate+0x4b/0x60
 [<ffffffff8100a69d>] kvm_irqfd+0x39d/0x580
 [<ffffffff81007a27>] kvm_vm_ioctl+0x207/0x5b0
 [<ffffffff810c9545>] ? update_curr+0xf5/0x180
 [<ffffffff811b66e8>] do_vfs_ioctl+0x98/0x550
 [<ffffffff810c1f5e>] ? finish_task_switch+0x4e/0xe0
 [<ffffffff81c054aa>] ? __schedule+0x2ea/0x710
 [<ffffffff811b6bf7>] sys_ioctl+0x57/0x90
 [<ffffffff8140ae9e>] ? trace_hardirqs_on_thunk+0x3a/0x3c
 [<ffffffff81c0f602>] system_call_fastpath+0x16/0x1b
Code: c1 ea 08 38 c2 74 0f 66 0f 1f 44 00 00 f3 90 0f b6 03 38 c2 75 f7 48 83 c4 08 5b c9 c3 55 48 89 e5 66 66 66 66 90 b8 00 01 00 00 <f0> 66 0f c1 07 89 c2 66 c1 ea 08 38 c2 74 0c 0f 1f 00 f3 90 0f
RIP  [<ffffffff81c0721e>] _raw_spin_lock+0xe/0x30
RSP <ffff880221721cc8>
CR2: 0000000000000000
---[ end trace 13fb1e4b6e5ab21f ]---

Signed-off-by: Asias He <asias@redhat.com>
Acked-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>

diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
index 45f09362ee7be02df67171efa3508fa225129499..8a1889ccb8830174339debe1e16cd353141af225 100644 (file)
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -3105,13 +3105,21 @@ int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
        int r;
        int cpu;
 
-       r = kvm_irqfd_init();
-       if (r)
-               goto out_irqfd;
        r = kvm_arch_init(opaque);
        if (r)
                goto out_fail;
 
+       /*
+        * kvm_arch_init makes sure there's at most one caller
+        * for architectures that support multiple implementations,
+        * like intel and amd on x86.
+        * kvm_arch_init must be called before kvm_irqfd_init to avoid creating
+        * conflicts in case kvm is already setup for another implementation.
+        */
+       r = kvm_irqfd_init();
+       if (r)
+               goto out_irqfd;
+
        if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) {
                r = -ENOMEM;
                goto out_free_0;
@@ -3186,10 +3194,10 @@ out_free_1:
 out_free_0a:
        free_cpumask_var(cpus_hardware_enabled);
 out_free_0:
-       kvm_arch_exit();
-out_fail:
        kvm_irqfd_exit();
 out_irqfd:
+       kvm_arch_exit();
+out_fail:
        return r;
 }
 EXPORT_SYMBOL_GPL(kvm_init);