Skip to contentSkip to conversation

rfkill: prep for rfkill API changes

PATCH In review Johannes Berg 3 participants 5 messages Jul 5, 2009 – Jul 6, 2009 Rfkill

lore
Share & tools
lore

Johannes Berg

author maintainer
#0 lore raw

PATCH 1/1 rfkill: prep for rfkill API changes

We've designed the /dev/rfkill API in a way that we can increase the event struct by adding members at the end, should it become necessary. To validate the events, userspace and the kernel need to have the proper event size to check for -- when reading from the other end they need to verify that it's at least version 1 of the event API, with the current struct size, so define a constant for that and make the code a little more 'future proof'.

Not that I expect that we'll have to change the event size any time soon, but it's better to write the code in a way that lends itself to extending.

Due to the current size of the event struct, the code is currently equivalent, but should the event struct ever need to be increased the new code might not need changing.

---
 include/linux/rfkill.h |   14 ++++++++++++++
 net/rfkill/core.c      |   10 ++++++++--
 2 files changed, 22 insertions(+), 2 deletions(-)
include/linux/rfkill.h +14 -0
@@ -82,6 +82,20 @@ struct rfkill_event {
 	__u8  soft, hard;
 } __packed;
 
+/*
+ * We are planning to be backward and forward compatible with changes
+ * to the event struct, by adding new, optional, members at the end.
+ * When reading an event (whether the kernel from userspace or vice
+ * versa) we need to accept anything that's at least as large as the
+ * version 1 event size, but might be able to accept other sizes in
+ * the future.
+ *
+ * One exception is the kernel -- we already have two event sizes in
+ * that we've made the 'hard' member optional since our only option
+ * is to ignore it anyway.
+ */
+#define RFKILL_EVENT_SIZE_V1	8
+
 /* ioctl for turning off rfkill-input (if present) */
 #define RFKILL_IOC_MAGIC	'R'
 #define RFKILL_IOC_NOINPUT	1
net/rfkill/core.c +8 -2
@@ -1076,10 +1076,16 @@ static ssize_t rfkill_fop_write(struct f
 	struct rfkill_event ev;
 
 	/* we don't need the 'hard' variable but accept it */
-	if (count < sizeof(ev) - 1)
+	if (count < RFKILL_EVENT_SIZE_V1 - 1)
 		return -EINVAL;
 
-	if (copy_from_user(&ev, buf, sizeof(ev) - 1))
+	/*
+	 * Copy as much data as we can accept into our 'ev' buffer,
+	 * but tell userspace how much we've copied so it can determine
+	 * our API version even in a write() call, if it cares.
+	 */
+	count = min(count, sizeof(ev));
+	if (copy_from_user(&ev, buf, count))
 		return -EFAULT;
 
 	if (ev.op != RFKILL_OP_CHANGE && ev.op != RFKILL_OP_CHANGE_ALL)
#1 lore raw
Quoted 10 lines · Johannes Berg

We've designed the /dev/rfkill API in a way that we can increase the event struct by adding members at the end, should it become necessary. To validate the events, userspace and the kernel need to have the proper event size to check for -- when reading from the other end they need to verify that it's at least version 1 of the event API, with the current struct size, so define a constant for that and make the code a little more 'future proof'.

Quoted 8 lines

Not that I expect that we'll have to change the event size any time soon, but it's better to write the code in a way that lends itself to extending.

Due to the current size of the event struct, the code is currently equivalent, but should the event struct ever need to be increased the new code might not need changing.

This is good, but not complete. Changes to the API that don't change the size (or those which reduces it) are a problem: they are impossible with a schema that detects API version by the size alone.

Maybe a new field with a API serial number should be added right _now_ while it is still not too painful to do it? This would be V2 of the API (detectable by the size change), but there aren't many users yet, so the effort to support it would not be daunting.

You can also publish the API version through a read-only sysfs attribute or a separate IOCTL... it doesn't really need to be in-band (although in-band is a lot better).

Signature · 4 lines
  "One disk to rule them all, One disk to find them. One disk to bring
  them all and in the darkness grind them. In the Land of Redmond
  where the shadows lie." -- The Silicon Valley Tarot
  Henrique Holschuh
author maintainer
#2 lore raw
Quoted 3 lines · Henrique de Moraes Holschuh

This is good, but not complete. Changes to the API that don't change the size (or those which reduces it) are a problem: they are impossible with a schema that detects API version by the size alone.

Umm, no, we can also add operations as required already.

Quoted 8 lines

Maybe a new field with a API serial number should be added right _now_ while it is still not too painful to do it? This would be V2 of the API (detectable by the size change), but there aren't many users yet, so the effort to support it would not be daunting.

You can also publish the API version through a read-only sysfs attribute or a separate IOCTL... it doesn't really need to be in-band (although in-band is a lot better).

No.

This patch doesn't change anything at all. It just reduces the _subtlety_ involved in actually doing a size change, nothing more. A real event version is _not_ necessary at all. In fact, I would argue that basing _anything_ on the "version" rather than the "feature set" is useless.

johannes

#3 lore raw
Quoted 6 lines · Johannes Berg
Quoted 3 lines · Henrique de Moraes Holschuh

This is good, but not complete. Changes to the API that don't change the size (or those which reduces it) are a problem: they are impossible with a schema that detects API version by the size alone.

Umm, no, we can also add operations as required already.

Yeah, I know. What we can't do is to change the current ones, or drop them.

Quoted 4 lines

_subtlety_ involved in actually doing a size change, nothing more. A real event version is _not_ necessary at all. In fact, I would argue that basing _anything_ on the "version" rather than the "feature set" is useless.

IMO version can tell you the supported feature set, but one could use a feature set bitmap as well if one wanted. Anyway, if you guys deem the current way good enough, I am fine with it.

Signature · 4 lines
  "One disk to rule them all, One disk to find them. One disk to bring
  them all and in the darkness grind them. In the Land of Redmond
  where the shadows lie." -- The Silicon Valley Tarot
  Henrique Holschuh
Acked-by on v1 1/1
#4 lore raw
In reply to Johannes Berg

Hi Johannes,

Quoted 14 lines

We've designed the /dev/rfkill API in a way that we can increase the event struct by adding members at the end, should it become necessary. To validate the events, userspace and the kernel need to have the proper event size to check for -- when reading from the other end they need to verify that it's at least version 1 of the event API, with the current struct size, so define a constant for that and make the code a little more 'future proof'.

Not that I expect that we'll have to change the event size any time soon, but it's better to write the code in a way that lends itself to extending.

Quoted 6 lines

Due to the current size of the event struct, the code is currently equivalent, but should the event struct ever need to be increased the new code might not need changing.

patch is fine as it is. We have discussed future extensions in this area and I don't see any problems.

Regards

Marcel