On Mon, Dec 25, 2017 at 12:28:39PM +0200, Leon Romanovsky wrote:
On Fri, Dec 22, 2017 at 01:38:52PM -0800, Bart Van Assche wrote:
Avoid that the following two warnings are reported when registering an rxe device:
WARNING: CPU: 2 PID: 1005 at drivers/infiniband/core/device.c:449 ib_register_device+0x591/0x640 [ib_core] CPU: 2 PID: 1005 Comm: run_tests Not tainted 4.15.0-rc4-dbg+ #2 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.0.0-prebuilt.qemu-project.org 04/01/2014 RIP: 0010:ib_register_device+0x591/0x640 [ib_core] Call Trace: rxe_register_device+0x3c6/0x470 [rdma_rxe] rxe_add+0x543/0x5e0 [rdma_rxe] rxe_net_add+0x37/0xb0 [rdma_rxe] rxe_param_set_add+0x5a/0x120 [rdma_rxe] param_attr_store+0x5e/0xc0 module_attr_store+0x19/0x30 sysfs_kf_write+0x3d/0x50 kernfs_fop_write+0x116/0x1a0 __vfs_write+0x23/0x120 vfs_write+0xbe/0x1b0 SyS_write+0x44/0xa0 entry_SYSCALL_64_fastpath+0x23/0x9a
WARNING: CPU: 2 PID: 1005 at drivers/infiniband/core/sysfs.c:1279 ib_device_register_sysfs+0x11d/0x160 [ib_core] CPU: 2 PID: 1005 Comm: run_tests Tainted: G W 4.15.0-rc4-dbg+ #2 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.0.0-prebuilt.qemu-project.org 04/01/2014 RIP: 0010:ib_device_register_sysfs+0x11d/0x160 [ib_core] Call Trace: ib_register_device+0x3f7/0x640 [ib_core] rxe_register_device+0x3c6/0x470 [rdma_rxe] rxe_add+0x543/0x5e0 [rdma_rxe] rxe_net_add+0x37/0xb0 [rdma_rxe] rxe_param_set_add+0x5a/0x120 [rdma_rxe] param_attr_store+0x5e/0xc0 module_attr_store+0x19/0x30 sysfs_kf_write+0x3d/0x50 kernfs_fop_write+0x116/0x1a0 __vfs_write+0x23/0x120 vfs_write+0xbe/0x1b0 SyS_write+0x44/0xa0 entry_SYSCALL_64_fastpath+0x23/0x9a
Fixes: commit 99db9494035f ("IB/core: Remove ib_device.dma_device")
^^^^^^ redundant
Signed-off-by: Bart Van Assche bart.vanassche@wdc.com Cc: Leon Romanovsky leon@kernel.org Cc: stable@vger.kernel.org # v4.11 drivers/infiniband/core/device.c | 2 +- drivers/infiniband/core/sysfs.c | 1 - 2 files changed, 1 insertion(+), 2 deletions(-)
diff --git a/drivers/infiniband/core/device.c b/drivers/infiniband/core/device.c index 30914f3baa5f..b9d7570e9989 100644 +++ b/drivers/infiniband/core/device.c @@ -446,7 +446,6 @@ int ib_register_device(struct ib_device *device, struct ib_udata uhw = {.outlen = 0, .inlen = 0}; struct device *parent = device->dev.parent;
- WARN_ON_ONCE(!parent); WARN_ON_ONCE(device->dma_device); if (device->dev.dma_ops) {
I understand that it fixed warning for you, but why is it correct? You are using "parent" in this "if (device->dev.dma_ops)" too.
If no parent is specified then the caller must provide a full set of DMA data and none should be copied from the parent.
Presumably rxe does this, and has no parent because it is a virtual device.
I'm a little more worried about the ib_device_register_sysfs - why was the WARN_ON ever there?
What do you think about this as a clarification Bart?
diff --git a/drivers/infiniband/core/device.c b/drivers/infiniband/core/device.c index bb2686d56d3c65..c535c7803d80a7 100644 --- a/drivers/infiniband/core/device.c +++ b/drivers/infiniband/core/device.c @@ -446,7 +446,6 @@ int ib_register_device(struct ib_device *device, struct ib_udata uhw = {.outlen = 0, .inlen = 0}; struct device *parent = device->dev.parent;
- WARN_ON_ONCE(!parent); WARN_ON_ONCE(device->dma_device); if (device->dev.dma_ops) { /* @@ -455,16 +454,20 @@ int ib_register_device(struct ib_device *device, * into device->dev. */ device->dma_device = &device->dev; - if (!device->dev.dma_mask) - device->dev.dma_mask = parent->dma_mask; - if (!device->dev.coherent_dma_mask) - device->dev.coherent_dma_mask = - parent->coherent_dma_mask; + + if (parent) { + if (!device->dev.dma_mask) + device->dev.dma_mask = parent->dma_mask; + if (!device->dev.coherent_dma_mask) + device->dev.coherent_dma_mask = + parent->coherent_dma_mask; + } } else { /* * The caller did not provide custom DMA operations. Use the * DMA mapping operations of the parent device. */ + WARN_ON_ONCE(!parent); device->dma_device = parent; }
On Wed, 2017-12-27 at 15:56 -0700, Jason Gunthorpe wrote:
I'm a little more worried about the ib_device_register_sysfs - why was the WARN_ON ever there?
That WARN_ON() statement was introduced by commit 97a9ea848016 ("IB/core: Initialize ib_device.dev.parent earlier"):
[ ... ] diff --git a/drivers/infiniband/core/sysfs.c b/drivers/infiniband/core/sysfs.c index c1fb545e8d78..daadf3130c9f 100644 --- a/drivers/infiniband/core/sysfs.c +++ b/drivers/infiniband/core/sysfs.c @@ -1258,7 +1258,7 @@ int ib_device_register_sysfs(struct ib_device *device, int ret; int i;
- device->dev.parent = device->dma_device; + WARN_ON_ONCE(!device->dev.parent); ret = dev_set_name(class_dev, "%s", device->name); if (ret) return ret; [ ... ]
Since ib_device_register_sysfs() does not use the parent pointer in any way I think it is sufficient to check the parent pointer in ib_register_device() and to leave out any parent pointer checks from ib_device_register_sysfs().
What do you think about this as a clarification Bart? [ ... ]
OK, I will add a parent pointer check in the code paths that dereference that pointer.
Bart.
linux-stable-mirror@lists.linaro.org