Le 09/01/2019 à 17:25, Greg Kurz a écrit :
On Wed, 9 Jan 2019 16:13:42 +0100
Frederic Barrat <fbarrat@xxxxxxxxxxxxx> wrote:
With a recent change around IOMMU group, a system with an opencapi
adapter is no longer booting and we get a kernel oops:
BUG: Kernel NULL pointer dereference at 0x00000028
Faulting instruction address: 0xc0000000000aa38c
Oops: Kernel access of bad area, sig: 7 [#1]
LE SMP NR_CPUS=2048 NUMA PowerNV
Modules linked in:
CPU: 5 PID: 1 Comm: swapper/4 Not tainted 5.0.0-rc1-fxb-00001-g3bd6e94bec12
NIP: c0000000000aa38c LR: c0000000000a6608 CTR: c000000000097480
REGS: c000000005783700 TRAP: 0300 Not tainted (5.0.0-rc1-fxb-00001-g3bd6
MSR: 9000000002009033 <SF,HV,VEC,EE,ME,IR,DR,RI,LE> CR: 28000228 XER: 20
CFAR: c0000000000a6604 DAR: 0000000000000028 DSISR: 00080000 IRQMASK: 0
GPR00: c0000000000a6608 c000000005783990 c000000001036100 c0000007bf761860
GPR04: 0000000000000000 c000000005783834 0000000000000000 0000000000000000
GPR08: 69626d2c6e707500 0000000000000000 0000000000000000 9000000002001003
GPR12: 0000000000000000 c0000007bfff8300 c000000000010450 0000000000000000
GPR16: c000000000ced938 0000000000000100 c000000000ced948 00000000000a0000
GPR20: 00000000000bfffe c000000000ced9a8 0000000000000200 c000000000ced978
GPR24: 00000000006080c0 c000000716d09828 c00000002e6fd000 0000000000000000
GPR28: c0000007bf4aff68 c0000007bf8d0080 c000000000f23938 c0000007bf761860
NIP [c0000000000aa38c] pnv_try_setup_npu_table_group+0x1c/0x1a0
LR [c0000000000a6608] pnv_pci_ioda_fixup+0x1f8/0x660
Call Trace:
[c000000005783990] [c0000000000aa3d0] pnv_try_setup_npu_table_group+0x60/0x
[c0000000057839d0] [c0000000000a661c] pnv_pci_ioda_fixup+0x20c/0x660
[c000000005783ab0] [c000000000e1d4c0] pcibios_resource_survey+0x2c8/0x31c
[c000000005783b90] [c000000000e1caf4] pcibios_init+0xb0/0xe4
[c000000005783c10] [c000000000010054] do_one_initcall+0x64/0x264
[c000000005783ce0] [c000000000e1132c] kernel_init_freeable+0x36c/0x468
[c000000005783db0] [c000000000010474] kernel_init+0x2c/0x148
[c000000005783e20] [c00000000000b794] ret_from_kernel_thread+0x5c/0x68
An opencapi device is using a device PE, so the current code breaks
because pe->pbus is not defined.
More generally, there's no need to define an IOMMU group for opencapi,
as the device sends real addresses directly (admittedly, the
virtualization story is yet to be written). So let's fix it by
Current plan is to go for mediated VFIO. The real HW stays under the control
of the host ocxl driver, and we still don't need an IOMMU group.
skipping the IOMMU group setup for opencapi PHBs.
Fixes: 0bd971676e68 ("powerpc/powernv/npu: Add compound IOMMU groups")
Signed-off-by: Frederic Barrat <fbarrat@xxxxxxxxxxxxx>
---
Reviewed-by: Greg Kurz <groug@xxxxxxxx>
and
Cc: stable@xxxxxxxxxxxxxxx # v4.20
Thanks for the review! But why did you add stable? that problem is only
seen on 5.0-rc1, isn't it?
Fred
arch/powerpc/platforms/powernv/pci-ioda.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/arch/powerpc/platforms/powernv/pci-ioda.c b/arch/powerpc/platforms/powernv/pci-ioda.c
index 1d6406a051f1..7db3119f8a5b 100644
--- a/arch/powerpc/platforms/powernv/pci-ioda.c
+++ b/arch/powerpc/platforms/powernv/pci-ioda.c
@@ -2681,7 +2681,8 @@ static void pnv_pci_ioda_setup_iommu_api(void)
list_for_each_entry(hose, &hose_list, list_node) {
phb = hose->private_data;
- if (phb->type == PNV_PHB_NPU_NVLINK)
+ if (phb->type == PNV_PHB_NPU_NVLINK ||
+ phb->type == PNV_PHB_NPU_OCAPI)
continue;
list_for_each_entry(pe, &phb->ioda.pe_list, list) {
