Skip to contentSkip to conversation

dax: Change return type to vm_fault_t

PATCH Superseded Souptick Joarder 3 participants 4 messages Apr 16, 2018 – Apr 17, 2018

Subsystems (2)

v2 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 dax: Change return type to vm_fault_t

Use new return type vm_fault_t for fault and huge_fault handler. For now, this is just documenting that the function returns a VM_FAULT value rather than an errno. Once all instances are converted, vm_fault_t will become a distinct type.

Reference id -> 1c8f422059ae ("mm: change return type to vm_fault_t")

Previously vm_insert_mixed() returns err which driver mapped into VM_FAULT_* type. The new function vmf_insert_mixed() will replace this inefficiency by returning VM_FAULT_* type.

---
v2: Modified the change log

v3: Updated the change log

 drivers/dax/device.c | 26 +++++++++++---------------
 1 file changed, 11 insertions(+), 15 deletions(-)
drivers/dax/device.c +11 -15
index 2137dbc..a122701 100644
@@ -243,11 +243,11 @@ __weak phys_addr_t dax_pgoff_to_phys(struct dev_dax *dev_dax, pgoff_t pgoff,
 	return -1;
 }
 
-static int __dev_dax_pte_fault(struct dev_dax *dev_dax, struct vm_fault *vmf)
+static vm_fault_t __dev_dax_pte_fault(struct dev_dax *dev_dax,
+				struct vm_fault *vmf)
 {
 	struct device *dev = &dev_dax->dev;
 	struct dax_region *dax_region;
-	int rc = VM_FAULT_SIGBUS;
 	phys_addr_t phys;
 	pfn_t pfn;
 	unsigned int fault_size = PAGE_SIZE;
@@ -274,17 +274,11 @@ static int __dev_dax_pte_fault(struct dev_dax *dev_dax, struct vm_fault *vmf)
 
 	pfn = phys_to_pfn_t(phys, dax_region->pfn_flags);
 
-	rc = vm_insert_mixed(vmf->vma, vmf->address, pfn);
-
-	if (rc == -ENOMEM)
-		return VM_FAULT_OOM;
-	if (rc < 0 && rc != -EBUSY)
-		return VM_FAULT_SIGBUS;
-
-	return VM_FAULT_NOPAGE;
+	return vmf_insert_mixed(vmf->vma, vmf->address, pfn);
 }
 
-static int __dev_dax_pmd_fault(struct dev_dax *dev_dax, struct vm_fault *vmf)
+static vm_fault_t __dev_dax_pmd_fault(struct dev_dax *dev_dax,
+				struct vm_fault *vmf)
 {
 	unsigned long pmd_addr = vmf->address & PMD_MASK;
 	struct device *dev = &dev_dax->dev;
@@ -335,7 +329,8 @@ static int __dev_dax_pmd_fault(struct dev_dax *dev_dax, struct vm_fault *vmf)
 }
 
 #ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
-static int __dev_dax_pud_fault(struct dev_dax *dev_dax, struct vm_fault *vmf)
+static vm_fault_t __dev_dax_pud_fault(struct dev_dax *dev_dax,
+				struct vm_fault *vmf)
 {
 	unsigned long pud_addr = vmf->address & PUD_MASK;
 	struct device *dev = &dev_dax->dev;
@@ -386,13 +381,14 @@ static int __dev_dax_pud_fault(struct dev_dax *dev_dax, struct vm_fault *vmf)
 			vmf->flags & FAULT_FLAG_WRITE);
 }
 #else
-static int __dev_dax_pud_fault(struct dev_dax *dev_dax, struct vm_fault *vmf)
+static vm_fault_t __dev_dax_pud_fault(struct dev_dax *dev_dax,
+				struct vm_fault *vmf)
 {
 	return VM_FAULT_FALLBACK;
 }
 #endif /* !CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD */
 
-static int dev_dax_huge_fault(struct vm_fault *vmf,
+static vm_fault_t dev_dax_huge_fault(struct vm_fault *vmf,
 		enum page_entry_size pe_size)
 {
 	int rc, id;
@@ -423,7 +419,7 @@ static int dev_dax_huge_fault(struct vm_fault *vmf,
 	return rc;
 }
 
-static int dev_dax_fault(struct vm_fault *vmf)
+static vm_fault_t dev_dax_fault(struct vm_fault *vmf)
 {
 	return dev_dax_huge_fault(vmf, PE_SIZE_PTE);
 }
Signature · 6 lines
1.9.1

_______________________________________________
Linux-nvdimm mailing list
[email protected]
https://lists.01.org/mailman/listinfo/linux-nvdimm
Reviewed-by on v2 1/1
#1 lore raw
Quoted 5 lines · Souptick Joarder

Use new return type vm_fault_t for fault and huge_fault handler. For now, this is just documenting that the function returns a VM_FAULT value rather than an errno. Once all instances are converted, vm_fault_t will become a distinct type.

I'm not sure what you mean by "vm_fault_t will become a distinct type"? Do you mean you'll make it into an enum, i.e.:

enum vm_fault_t {
	VM_FAULT_OOM		= 0x0001,
	VM_FAULT_SIGBUS		= 0x0002,
	VM_FAULT_MAJOR		= 0x0004,
	VM_FAULT_WRITE		= 0x0008,
	VM_FAULT_HWPOISON	= 0x0010,
	VM_FAULT_HWPOISON_LARGE	= 0x0020,
	VM_FAULT_NOPAGE		= 0x0100,
	VM_FAULT_LOCKED		= 0x0200,
	VM_FAULT_RETRY		= 0x0400,
	VM_FAULT_FALLBACK	= 0x0800,
	VM_FAULT_DONE_COW	= 0x1000,
	VM_FAULT_NEEDDSYNC	= 0x2000,
};

If so, I think that would be great for readability. Right now I look up the definition of vm_fault_t and see it's typedef'd to an int, and that doesn't give me as much info as the above enum would.

If this is the plan, I don't understand why you need to make all the site conversions first?

Quoted 8 lines

Reference id -> 1c8f422059ae ("mm: change return type to vm_fault_t")

Previously vm_insert_mixed() returns err which driver mapped into VM_FAULT_* type. The new function vmf_insert_mixed() will replace this inefficiency by returning VM_FAULT_* type.

Quoted 6 lines

---
v2: Modified the change log

v3: Updated the change log

You've only got a v2. :)

Sure, the code looks correct. You can add:

_______________________________________________
Linux-nvdimm mailing list
[email protected]
https://lists.01.org/mailman/listinfo/linux-nvdimm

#2 lore raw
Quoted 17 lines · Ross Zwisler

I'm not sure what you mean by "vm_fault_t will become a distinct type"? Do you mean you'll make it into an enum, i.e.:

enum vm_fault_t {
	VM_FAULT_OOM		= 0x0001,
	VM_FAULT_SIGBUS		= 0x0002,
	VM_FAULT_MAJOR		= 0x0004,
	VM_FAULT_WRITE		= 0x0008,
	VM_FAULT_HWPOISON	= 0x0010,
	VM_FAULT_HWPOISON_LARGE	= 0x0020,
	VM_FAULT_NOPAGE		= 0x0100,
	VM_FAULT_LOCKED		= 0x0200,
	VM_FAULT_RETRY		= 0x0400,
	VM_FAULT_FALLBACK	= 0x0800,
	VM_FAULT_DONE_COW	= 0x1000,
	VM_FAULT_NEEDDSYNC	= 0x2000,
};

My current tree has this in it:

+/*
+ * Different kinds of faults, as returned from fault handlers.
+ * Used to decide whether a process gets delivered SIGBUS or
+ * just gets major/minor fault counters bumped up.
+ */
+typedef __bitwise unsigned int vm_fault_t;
+enum {

+       VM_FAULT_OOM            = (__force vm_fault_t)0x000001, /* Out Of Memory */
+       VM_FAULT_SIGBUS         = (__force vm_fault_t)0x000002, /* Bad access */
+       VM_FAULT_MAJOR          = (__force vm_fault_t)0x000004, /* Page read from storage */
+       VM_FAULT_WRITE          = (__force vm_fault_t)0x000008, /* Special case for get_user_pages */
+       VM_FAULT_HWPOISON       = (__force vm_fault_t)0x000010, /* Hit poisoned small page */
+       VM_FAULT_HWPOISON_LARGE = (__force vm_fault_t)0x000020, /* Hit poisoned large page. Index encoded in upper bits */
+       VM_FAULT_SIGSEGV        = (__force vm_fault_t)0x000040,
+       VM_FAULT_NOPAGE         = (__force vm_fault_t)0x000100, /* ->fault installed the pte, not return page */
+       VM_FAULT_LOCKED         = (__force vm_fault_t)0x000200, /* ->fault locked the returned page */
+       VM_FAULT_RETRY          = (__force vm_fault_t)0x000400, /* ->fault blocked, must retry */
+       VM_FAULT_FALLBACK       = (__force vm_fault_t)0x000800, /* huge page fault failed, fall back to small */
+       VM_FAULT_DONE_COW       = (__force vm_fault_t)0x001000, /* ->fault has fully handled COW */
+       VM_FAULT_NEEDDSYNC      = (__force vm_fault_t)0x002000, /* ->fault did not modify page tables

+ * and needs fsync() to complete (for
+ * synchronous page faults in DAX) */
+ VM_FAULT_HINDEX_MASK = (__force vm_fault_t)0x0f0000,
+};

However, I'm hoping to get a better syntax into sparse ;-)

Quoted 6 lines

If so, I think that would be great for readability. Right now I look up the definition of vm_fault_t and see it's typedef'd to an int, and that doesn't give me as much info as the above enum would.

If this is the plan, I don't understand why you need to make all the site conversions first?

Changing it from a signed to an unsigned int causes GCC to warn about an assignment from an incompatible type -- int foo(void) is incompatible with unsigned int foo(void).

Quoted 2 lines

Sure, the code looks correct. You can add:

Thanks, Ross.
_______________________________________________
Linux-nvdimm mailing list
[email protected]
https://lists.01.org/mailman/listinfo/linux-nvdimm

#3 lore raw
In reply to Ross Zwisler
Quoted 40 lines · Ross Zwisler
Quoted 5 lines · Souptick Joarder

Use new return type vm_fault_t for fault and huge_fault handler. For now, this is just documenting that the function returns a VM_FAULT value rather than an errno. Once all instances are converted, vm_fault_t will become a distinct type.

I'm not sure what you mean by "vm_fault_t will become a distinct type"? Do you mean you'll make it into an enum, i.e.:

enum vm_fault_t {
        VM_FAULT_OOM            = 0x0001,
        VM_FAULT_SIGBUS         = 0x0002,
        VM_FAULT_MAJOR          = 0x0004,
        VM_FAULT_WRITE          = 0x0008,
        VM_FAULT_HWPOISON       = 0x0010,
        VM_FAULT_HWPOISON_LARGE = 0x0020,
        VM_FAULT_NOPAGE         = 0x0100,
        VM_FAULT_LOCKED         = 0x0200,
        VM_FAULT_RETRY          = 0x0400,
        VM_FAULT_FALLBACK       = 0x0800,
        VM_FAULT_DONE_COW       = 0x1000,
        VM_FAULT_NEEDDSYNC      = 0x2000,
};

If so, I think that would be great for readability. Right now I look up the definition of vm_fault_t and see it's typedef'd to an int, and that doesn't give me as much info as the above enum would.

If this is the plan, I don't understand why you need to make all the site conversions first?

Quoted 8 lines

Reference id -> 1c8f422059ae ("mm: change return type to vm_fault_t")

Previously vm_insert_mixed() returns err which driver mapped into VM_FAULT_* type. The new function vmf_insert_mixed() will replace this inefficiency by returning VM_FAULT_* type.

Quoted 8 lines
Quoted 6 lines

---
v2: Modified the change log

v3: Updated the change log

You've only got a v2. :)

Sorry about it. I will correct :)

Quoted 3 lines

Sure, the code looks correct. You can add:

Thanks.
_______________________________________________
Linux-nvdimm mailing list
[email protected]
https://lists.01.org/mailman/listinfo/linux-nvdimm