[PATCH] device-dax: switch to srcu, fix rcu_read_lock() vs pte allocation

[Dan Williams]

The following warning triggers with a new unit test that stresses the
device-dax interface.

 ===============================
 [ ERR: suspicious RCU usage.  ]
 4.11.0-rc4+ #1049 Tainted: G           O
 -------------------------------
 ./include/linux/rcupdate.h:521 Illegal context switch in RCU read-side critical section!

 other info that might help us debug this:

 rcu_scheduler_active = 2, debug_locks = 0
 2 locks held by fio/9070:
  #0:  (&mm->mmap_sem){++++++}, at: [<ffffffff8d0739d7>] __do_page_fault+0x167/0x4f0
  #1:  (rcu_read_lock){......}, at: [<ffffffffc03fbd02>] dax_dev_huge_fault+0x32/0x620 [dax]

 Call Trace:
  dump_stack+0x86/0xc3
  lockdep_rcu_suspicious+0xd7/0x110
  ___might_sleep+0xac/0x250
  __might_sleep+0x4a/0x80
  __alloc_pages_nodemask+0x23a/0x360
  alloc_pages_current+0xa1/0x1f0
  pte_alloc_one+0x17/0x80
  __pte_alloc+0x1e/0x120
  __get_locked_pte+0x1bf/0x1d0
  insert_pfn.isra.70+0x3a/0x100
  ? lookup_memtype+0xa6/0xd0
  vm_insert_mixed+0x64/0x90
  dax_dev_huge_fault+0x520/0x620 [dax]
  ? dax_dev_huge_fault+0x32/0x620 [dax]
  dax_dev_fault+0x10/0x20 [dax]
  __do_fault+0x1e/0x140
  __handle_mm_fault+0x9af/0x10d0
  handle_mm_fault+0x16d/0x370
  ? handle_mm_fault+0x47/0x370
  __do_page_fault+0x28c/0x4f0
  trace_do_page_fault+0x58/0x2a0
  do_async_page_fault+0x1a/0xa0
  async_page_fault+0x28/0x30

Inserting a page table entry may trigger an allocation while we are
holding a read lock to keep the device instance alive for the duration
of the fault. Use srcu for this keep-alive protection.

Fixes: dee410792419 ("/dev/dax, core: file operations and dax-mmap")
Cc: <stable@xxxxxxxxxxxxxxx>
Signed-off-by: Dan Williams <dan.j.williams@xxxxxxxxx>
---
 drivers/dax/Kconfig |    1 +
 drivers/dax/dax.c   |   13 +++++++------
 2 files changed, 8 insertions(+), 6 deletions(-)

diff --git a/drivers/dax/Kconfig b/drivers/dax/Kconfig
index 3e2ab3b14eea..9e95bf94eb13 100644
--- a/drivers/dax/Kconfig
+++ b/drivers/dax/Kconfig
@@ -2,6 +2,7 @@ menuconfig DEV_DAX
 	tristate "DAX: direct access to differentiated memory"
 	default m if NVDIMM_DAX
 	depends on TRANSPARENT_HUGEPAGE
+	select SRCU
 	help
 	  Support raw access to differentiated (persistence, bandwidth,
 	  latency...) memory via an mmap(2) capable character
diff --git a/drivers/dax/dax.c b/drivers/dax/dax.c
index 80c6db279ae1..806f180c80d8 100644
--- a/drivers/dax/dax.c
+++ b/drivers/dax/dax.c
@@ -25,6 +25,7 @@
 #include "dax.h"
 
 static dev_t dax_devt;
+DEFINE_STATIC_SRCU(dax_srcu);
 static struct class *dax_class;
 static DEFINE_IDA(dax_minor_ida);
 static int nr_dax = CONFIG_NR_DEV_DAX;
@@ -60,7 +61,7 @@ struct dax_region {
  * @region - parent region
  * @dev - device backing the character device
  * @cdev - core chardev data
- * @alive - !alive + rcu grace period == no new mappings can be established
+ * @alive - !alive + srcu grace period == no new mappings can be established
  * @id - child id in the region
  * @num_resources - number of physical address extents in this device
  * @res - array of physical address ranges
@@ -569,7 +570,7 @@ static int __dax_dev_pud_fault(struct dax_dev *dax_dev, struct vm_fault *vmf)
 static int dax_dev_huge_fault(struct vm_fault *vmf,
 		enum page_entry_size pe_size)
 {
-	int rc;
+	int rc, id;
 	struct file *filp = vmf->vma->vm_file;
 	struct dax_dev *dax_dev = filp->private_data;
 
@@ -578,7 +579,7 @@ static int dax_dev_huge_fault(struct vm_fault *vmf,
 			? "write" : "read",
 			vmf->vma->vm_start, vmf->vma->vm_end);
 
-	rcu_read_lock();
+	id = srcu_read_lock(&dax_srcu);
 	switch (pe_size) {
 	case PE_SIZE_PTE:
 		rc = __dax_dev_pte_fault(dax_dev, vmf);
@@ -592,7 +593,7 @@ static int dax_dev_huge_fault(struct vm_fault *vmf,
 	default:
 		return VM_FAULT_FALLBACK;
 	}
-	rcu_read_unlock();
+	srcu_read_unlock(&dax_srcu, id);
 
 	return rc;
 }
@@ -713,11 +714,11 @@ static void unregister_dax_dev(void *dev)
 	 * Note, rcu is not protecting the liveness of dax_dev, rcu is
 	 * ensuring that any fault handlers that might have seen
 	 * dax_dev->alive == true, have completed.  Any fault handlers
-	 * that start after synchronize_rcu() has started will abort
+	 * that start after synchronize_srcu() has started will abort
 	 * upon seeing dax_dev->alive == false.
 	 */
 	dax_dev->alive = false;
-	synchronize_rcu();
+	synchronize_srcu(&dax_srcu);
 	unmap_mapping_range(dax_dev->inode->i_mapping, 0, 0, 1);
 	cdev_del(cdev);
 	device_unregister(dev);