On 27/11/2013 20:53, Norbert Marx wrote:
Hello,

on a Supermicro H8DGU server with the latest BIOS PCI passtrough fails.

I tried to give a two igbvf devices to the domU. pciback is configured and everything locks good until igbvf tries to initialize the PCI device. I have the same error with XEN 4.3, 4.3.1 and the current 4.4-unstable, linux kernel 3.12.0, 3.12.1, 3.9.

What about dom0 kernel?  This looks suspiciously like the new memory protection for MSI-X config tables.  With this model, pciback marks the MSI region as read only after it has set it up appropriately, with the knowledge that pcifront should indirect all requests.

What about Xen 4.2?

~Andrew


DomU config:
kernel = "/boot/gentoo-DomU"
memory = 2048
name   = "domU"
nic    = 2
vcpus  = 1
#pci    = ['02:10.0,msitranslate=1,permissive=1'] <= same crash
#pci    = ['02:10.0,permissive=1'] <= same crash
pci    = ['02:10.0', '02:10.1']
vif    = ['bridge=xenbr1','bridge=xenbr2']
disk   = ['phy:/dev/loop4,xvda1,w', 'phy:/dev/loop5,xvda2,w', 'phy:/dev/loop6,xvda3,w']
root   = "/dev/xvda1 ro rootfstype=ext4 iommu=soft xen-pcifront.verbose_request=1"

DomU crash log:

[   71.124852] pcifront pci-0: write dev=0000:00:00.0 - offset 72 size 2 val c00
2
[   71.124888] BUG: unable to handle kernel paging request at ffffc9000015400c
[   71.124900] IP: [<ffffffff8121ea05>] __msix_mask_irq+0x21/0x24
[   71.124911] PGD 784a0067 PUD 784a1067 PMD 784a2067 PTE 8010000000000464
[   71.124919] Oops: 0003 [#1] SMP
[   71.124923] Modules linked in: igbvf(+)
[   71.124929] CPU: 0 PID: 2114 Comm: insmod Not tainted 3.12.1-gentoo-DomU #6
[   71.124934] task: ffff8800784c3080 ti: ffff880077324000 task.ti: ffff88007732
4000
[   71.124939] RIP: e030:[<ffffffff8121ea05>]  [<ffffffff8121ea05>] __msix_mask_
irq+0x21/0x24
[   71.124947] RSP: e02b:ffff880077325bb0  EFLAGS: 00010286
[   71.124951] RAX: 0000000000000001 RBX: ffff880078741000 RCX: 0000000000000001
[   71.124957] RDX: ffffc9000015400c RSI: 0000000000000001 RDI: ffff880077770180
[   71.124961] RBP: ffff880077770180 R08: 0000000000000200 R09: ffff88007873fc00
[   71.124967] R10: 0000000000000000 R11: ffff88007873fc00 R12: 0000000000000000
[   71.124972] R13: ffff8800771138a0 R14: 0000000000000000 R15: ffffc9000015400c
[   71.124980] FS:  00007fb8d2c37700(0000) GS:ffff88007f200000(0000) knlGS:00000
00000000000
[   71.124985] CS:  e033 DS: 0000 ES: 0000 CR0: 000000008005003b
[   71.124989] CR2: ffffc9000015400c CR3: 000000007712e000 CR4: 0000000000040660
[   71.124994] Stack:
[   71.124996]  ffffffff8121f9dd ffff88007707c000 ffff880078741840 0000000000000
002
[   71.125004]  ffff880078741840 ffffffffa0005140 00000000c0020010 ffff880078741
000
[   71.125011]  ffff880078741000 0000000000000000 ffff880078741098 ffff88007707c
7c0
[   71.125018] Call Trace:
[   71.125023]  [<ffffffff8121f9dd>] ? pci_enable_msix+0x27d/0x353
[   71.125032]  [<ffffffffa00017b6>] ? igbvf_probe+0x323/0x8d9 [igbvf]
[   71.125039]  [<ffffffff8141fe9d>] ? _raw_spin_unlock_irqrestore+0x42/0x5b
[   71.125047]  [<ffffffff8121300d>] ? pci_device_probe+0x60/0x9d
[   71.125056]  [<ffffffff812af74d>] ? driver_probe_device+0x1b3/0x1b3
[   71.125060]  [<ffffffff812af62c>] ? driver_probe_device+0x92/0x1b3
[   71.125060]  [<ffffffff812af7a0>] ? __driver_attach+0x53/0x73
[   71.125060]  [<ffffffff812add94>] ? bus_for_each_dev+0x4e/0x7f
[   71.125060]  [<ffffffff812aedf2>] ? bus_add_driver+0xe5/0x22d
[   71.125060]  [<ffffffff812afcfa>] ? driver_register+0x82/0xb5
[   71.125060]  [<ffffffffa0008000>] ? 0xffffffffa0007fff
[   71.125060]  [<ffffffff81002092>] ? do_one_initcall+0x78/0x102
[   71.125060]  [<ffffffff810db633>] ? free_hot_cold_page+0x100/0x109
[   71.125060]  [<ffffffff811083b3>] ? kfree+0xb6/0xc8
[   71.125060]  [<ffffffff810fd8f4>] ? __vunmap+0x8c/0xc4
[   71.125060]  [<ffffffff810ad804>] ? load_module+0x18d3/0x1b9a
[   71.125060]  [<ffffffff810ab117>] ? mod_kobject_put+0x42/0x42
[   71.125060]  [<ffffffff81118716>] ? vfs_read+0xf7/0x13e
[   71.125060]  [<ffffffff810adbad>] ? SyS_finit_module+0x4e/0x62
[   71.125060]  [<ffffffff81420c8f>] ? tracesys+0xe1/0xe6
[   71.125060] Code: 83 c4 18 5b 5d 41 5c 41 5d c3 8b 47 08 0f b7 57 02 83 e0 fe
 c1 e2 04 89 c1 83 c9 01 83 c2 0c 85 f6 0f 45 c1 48 63 d2 48 03 57 28 <89> 02 c3
 48 8b 46 10 48 83 ef 48 48 85 c0 74 02 ff e0 48 c7 c0
[   71.125060] RIP  [<ffffffff8121ea05>] __msix_mask_irq+0x21/0x24
[   71.125060]  RSP <ffff880077325bb0>
[   71.125060] CR2: ffffc9000015400c
[   71.125060] ---[ end trace 66e59b16e50eead2 ]---

I tried also to use the patch from http://lists.xen.org/archives/html/xen-devel/2013-11/msg03752.html "[Xen-devel] [PATCH v6] x86: properly handle MSI-X unmask operation from guests" but without success.

More logs I attached. Any suggestions to fix this issue?

Kind regards,
Norbert


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel