Skip to contentSkip to conversation

ethtool: Add n-tuple string length to drvinfo and return it

PATCH Applied Jeff Kirsher 4 participants 12 messages Feb 26, 2010 – Feb 27, 2010

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ethtool: Add n-tuple string length to drvinfo and return it

From: Peter Waskiewicz <[email protected]>

The drvinfo struct should include the number of strings that get_rx_ntuple will return. It will be variable if an underlying driver implements its own get_rx_ntuple routine, so userspace needs to know how much data is coming.

---

 include/linux/ethtool.h |    1 +
 net/core/ethtool.c      |    3 +++
 2 files changed, 4 insertions(+), 0 deletions(-)
include/linux/ethtool.h +1 -0
index cca1c3d..f7992a2 100644
@@ -63,6 +63,7 @@ struct ethtool_drvinfo {
 	char	reserved2[12];
 	__u32	n_priv_flags;	/* number of flags valid in ETHTOOL_GPFLAGS */
 	__u32	n_stats;	/* number of u64's from ETHTOOL_GSTATS */
+	__u32	n_ntuples;	/* number of n-tuple filters from GSTRINGS */
 	__u32	testinfo_len;
 	__u32	eedump_len;	/* Size of data from ETHTOOL_GEEPROM (bytes) */
 	__u32	regdump_len;	/* Size of data from ETHTOOL_GREGS (bytes) */
net/core/ethtool.c +3 -0
index 31b1edd..1c94f48 100644
@@ -224,6 +224,9 @@ static noinline int ethtool_get_drvinfo(struct net_device *dev, void __user *use
 		rc = ops->get_sset_count(dev, ETH_SS_PRIV_FLAGS);
 		if (rc >= 0)
 			info.n_priv_flags = rc;
+		rc = ops->get_sset_count(dev, ETH_SS_NTUPLE_FILTERS);
+		if (rc >= 0)
+			info.n_ntuples = rc;
 	}
 	if (ops->get_regs_len)
 		info.regdump_len = ops->get_regs_len(dev);
maintainer Applied
#1 lore raw

From: Jeff Kirsher <[email protected]>
Date: Fri, 26 Feb 2010 03:54:20 -0800

Quoted 9 lines

From: Peter Waskiewicz <[email protected]>

The drvinfo struct should include the number of strings that get_rx_ntuple will return. It will be variable if an underlying driver implements its own get_rx_ntuple routine, so userspace needs to know how much data is coming.

Applied.

#2 lore raw
Quoted 10 lines · David Miller

From: Jeff Kirsher<[email protected]>
Date: Fri, 26 Feb 2010 03:54:20 -0800

Quoted 7 lines

From: Peter Waskiewicz<[email protected]>

The drvinfo struct should include the number of strings that get_rx_ntuple will return. It will be variable if an underlying driver implements its own get_rx_ntuple routine, so userspace needs to know how much data is coming.

Quoted 4 lines
Quoted 2 lines

Applied.

NAK. Did you even read the patch?

We don't increase the size of struct ethtool_drvinfo, _especially_ by sticking struct members into the middle of the struct.

What do you think 'reserved' is for???

Jeff

maintainer
#3 lore raw

From: Jeff Garzik <[email protected]>
Date: Fri, 26 Feb 2010 08:05:19 -0500

Quoted 1 line

NAK. Did you even read the patch?

I did, brain doesn't work sometimes :-)

Quoted 4 lines

We don't increase the size of struct ethtool_drvinfo, _especially_ by sticking struct members into the middle of the struct.

What do you think 'reserved' is for???

Right, I'll revert, thanks.

#4 lore raw
Quoted 10 lines · David Miller

From: Jeff Garzik <[email protected]>
Date: Fri, 26 Feb 2010 08:05:19 -0500

Quoted 1 line

NAK. Did you even read the patch?

I did, brain doesn't work sometimes :-)

Quoted 3 lines

We don't increase the size of struct ethtool_drvinfo, _especially_ by sticking struct members into the middle of the struct.

Brain lapse. When you mentioned ABI not being locked down until 2.6.34 is dropped, I didn't think.

I'm testing your proposed change, and once I have it working, I'll give it to Jeff K. to get pushed along with a new userspace patch.

Quoted 3 lines
Quoted 1 line

What do you think 'reserved' is for???

Right, I'll revert, thanks.

#5 lore raw
In reply to Jeff Kirsher
Quoted 7 lines · Jeff Kirsher

From: Peter Waskiewicz<[email protected]>

The drvinfo struct should include the number of strings that get_rx_ntuple will return. It will be variable if an underlying driver implements its own get_rx_ntuple routine, so userspace needs to know how much data is coming.

Quoted 7 lines

---

  include/linux/ethtool.h |    1 +
  net/core/ethtool.c      |    3 +++
  2 files changed, 4 insertions(+), 0 deletions(-)

(resending reply, standard patch-sending box is having trouble sending to vger)

As noted in the other email, your patch breaks ABI. The proper path is to decrease the size of reserved struct member, and NOT shift the offset of other members.

However, perhaps consider the following patch for returning n-tuple count, for four reasons:

1) space in ethtool_drvinfo is limited

2) the patch below permits trivial string set addition, without ABI changes beyond adding a new ETH_SS_xxx constant.

3) the patch below permits direct access to ops->get_sset_count(), rather than implicit access via ethtool_drvinfo

4) ethtool_drvinfo interface does not permit indication of ops->get_sset_count() failure, versus returning zero value. The patch below does so, via output sset_mask.

WARNING: this patch is compile-tested only.

NOTE: I added a cosmetic fix to ETHTOOL_[GS]RXNTUPLE constants, making their indentation consistent with the rest of the list of constants.

include/linux/ethtool.h +14 -3
index f7992a2..04ecf18 100644
@@ -254,6 +254,17 @@ struct ethtool_gstrings {
 	__u8	data[0];
 };
 
+struct ethtool_sset_info {
+	__u32	cmd;		/* ETHTOOL_GSSET_INFO */
+	__u32	reserved;
+	__u64	sset_mask;	/* input: each bit selects an sset to query */
+				/* output: each bit a returned sset */
+	__u32	data[0];	/* ETH_SS_xxx count, in order, based on bits
+				   in sset_mask.  One bit implies one
+				   __u32, two bits implies two
+				   __u32's, etc. */
+};
+
 enum ethtool_test_flags {
 	ETH_TEST_FL_OFFLINE	= (1 << 0),	/* online / offline */
 	ETH_TEST_FL_FAILED	= (1 << 1),	/* test passed / failed */
@@ -607,9 +618,9 @@ struct ethtool_ops {
 #define	ETHTOOL_SRXCLSRLINS	0x00000032 /* Insert RX classification rule */
 #define	ETHTOOL_FLASHDEV	0x00000033 /* Flash firmware to device */
 #define	ETHTOOL_RESET		0x00000034 /* Reset hardware */
-
-#define ETHTOOL_SRXNTUPLE	0x00000035 /* Add an n-tuple filter to device */
-#define ETHTOOL_GRXNTUPLE	0x00000036 /* Get n-tuple filters from device */
+#define	ETHTOOL_SRXNTUPLE	0x00000035 /* Add an n-tuple filter to device */
+#define	ETHTOOL_GRXNTUPLE	0x00000036 /* Get n-tuple filters from device */
+#define	ETHTOOL_GSSET_INFO	0x00000037 /* Get string set info */
 
 /* compatibility with older code */
 #define SPARC_ETH_GSET		ETHTOOL_GSET
net/core/ethtool.c +66 -0

78 lines not on this page · raw message

#6 lore raw
Quoted 15 lines · Jeff Garzik

However, perhaps consider the following patch for returning n-tuple count, for four reasons:

1) space in ethtool_drvinfo is limited

2) the patch below permits trivial string set addition, without ABI changes beyond adding a new ETH_SS_xxx constant.

3) the patch below permits direct access to ops->get_sset_count(), rather than implicit access via ethtool_drvinfo

4) ethtool_drvinfo interface does not permit indication of ops->get_sset_count() failure, versus returning zero value. The patch below does so, via output sset_mask.

Quoted 6 lines

WARNING: this patch is compile-tested only.

NOTE: I added a cosmetic fix to ETHTOOL_[GS]RXNTUPLE constants, making their indentation consistent with the rest of the list of constants.

With, perhaps, a note like the attached as a reminder to folks about future additions.

net/core/ethtool.c +3 -0
index 31b1edd..e376e8c 100644
@@ -212,6 +212,9 @@ static noinline int ethtool_get_drvinfo(struct net_device *dev, void __user *use
 	info.cmd = ETHTOOL_GDRVINFO;
 	ops->get_drvinfo(dev, &info);
 
+	/* this method of obtaining string set info is deprecated;
+	 * consider using ETHTOOL_GSSET_INFO instead
+	 */
 	if (ops->get_sset_count) {
 		int rc;
 
#7 lore raw
In reply to Jeff Garzik
Quoted 41 lines · Jeff Garzik
Quoted 7 lines · Jeff Kirsher

From: Peter Waskiewicz<[email protected]>

The drvinfo struct should include the number of strings that get_rx_ntuple will return. It will be variable if an underlying driver implements its own get_rx_ntuple routine, so userspace needs to know how much data is coming.

Quoted 7 lines

---

  include/linux/ethtool.h |    1 +
  net/core/ethtool.c      |    3 +++
  2 files changed, 4 insertions(+), 0 deletions(-)

(resending reply, standard patch-sending box is having trouble sending to vger)

As noted in the other email, your patch breaks ABI. The proper path is to decrease the size of reserved struct member, and NOT shift the offset of other members.

However, perhaps consider the following patch for returning n-tuple count, for four reasons:

1) space in ethtool_drvinfo is limited

2) the patch below permits trivial string set addition, without ABI changes beyond adding a new ETH_SS_xxx constant.

3) the patch below permits direct access to ops->get_sset_count(), rather than implicit access via ethtool_drvinfo

4) ethtool_drvinfo interface does not permit indication of ops->get_sset_count() failure, versus returning zero value. The patch below does so, via output sset_mask.

Quoted 6 lines

WARNING: this patch is compile-tested only.

NOTE: I added a cosmetic fix to ETHTOOL_[GS]RXNTUPLE constants, making their indentation consistent with the rest of the list of constants.

I'm updating your patch since I found an issue. The mask is passing in the ETH_SS_* flags, but then they're treated as bits, not enumerated flags. I'm thinking of the best non-intrusive way to correct it.

-PJ

#8 lore raw
Quoted 44 lines · Peter P Waskiewicz Jr
Quoted 43 lines · Jeff Garzik
Quoted 7 lines · Jeff Kirsher

From: Peter Waskiewicz<[email protected]>

The drvinfo struct should include the number of strings that get_rx_ntuple will return. It will be variable if an underlying driver implements its own get_rx_ntuple routine, so userspace needs to know how much data is coming.

Quoted 7 lines

---

   include/linux/ethtool.h |    1 +
   net/core/ethtool.c      |    3 +++
   2 files changed, 4 insertions(+), 0 deletions(-)

(resending reply, standard patch-sending box is having trouble sending to vger)

As noted in the other email, your patch breaks ABI. The proper path is to decrease the size of reserved struct member, and NOT shift the offset of other members.

However, perhaps consider the following patch for returning n-tuple count, for four reasons:

1) space in ethtool_drvinfo is limited

2) the patch below permits trivial string set addition, without ABI changes beyond adding a new ETH_SS_xxx constant.

3) the patch below permits direct access to ops->get_sset_count(), rather than implicit access via ethtool_drvinfo

4) ethtool_drvinfo interface does not permit indication of ops->get_sset_count() failure, versus returning zero value. The patch below does so, via output sset_mask.

WARNING: this patch is compile-tested only.

Quoted 8 lines
Quoted 4 lines

NOTE: I added a cosmetic fix to ETHTOOL_[GS]RXNTUPLE constants, making their indentation consistent with the rest of the list of constants.

I'm updating your patch since I found an issue. The mask is passing in the ETH_SS_* flags, but then they're treated as bits, not enumerated flags. I'm thinking of the best non-intrusive way to correct it.

As Ben noted, you cannot change those enumerated values, as they are already part of the ABI.

For ETHTOOL_GSSET_INFO, you initialize the sset_mask like this:

	info.sset_mask = (1ULL << ETH_SS_NTUPLE_FILTERS);

A multiple initialization would look like this:

	info.sset_mask = (1ULL << ETH_SS_NTUPLE_FILTERS) |
			(1ULL << ETH_SS_STATS) |
			(1ULL << ETH_SS_PRIV_FLAGS);

Do you still see an issue in my suggested code, now that sset_mask confusion is cleared up?

Regards,

Jeff

#9 lore raw
Quoted 59 lines · Jeff Garzik
Quoted 44 lines · Peter P Waskiewicz Jr
Quoted 43 lines · Jeff Garzik
Quoted 7 lines · Jeff Kirsher

From: Peter Waskiewicz<[email protected]>

The drvinfo struct should include the number of strings that get_rx_ntuple will return. It will be variable if an underlying driver implements its own get_rx_ntuple routine, so userspace needs to know how much data is coming.

Quoted 7 lines

---

include/linux/ethtool.h | 1 +
net/core/ethtool.c | 3 +++

2 files changed, 4 insertions(+), 0 deletions(-)

(resending reply, standard patch-sending box is having trouble sending to vger)

As noted in the other email, your patch breaks ABI. The proper path is to decrease the size of reserved struct member, and NOT shift the offset of other members.

However, perhaps consider the following patch for returning n-tuple count, for four reasons:

1) space in ethtool_drvinfo is limited

2) the patch below permits trivial string set addition, without ABI changes beyond adding a new ETH_SS_xxx constant.

3) the patch below permits direct access to ops->get_sset_count(), rather than implicit access via ethtool_drvinfo

4) ethtool_drvinfo interface does not permit indication of ops->get_sset_count() failure, versus returning zero value. The patch below does so, via output sset_mask.

WARNING: this patch is compile-tested only.

Quoted 8 lines
Quoted 4 lines

NOTE: I added a cosmetic fix to ETHTOOL_[GS]RXNTUPLE constants, making their indentation consistent with the rest of the list of constants.

I'm updating your patch since I found an issue. The mask is passing in the ETH_SS_* flags, but then they're treated as bits, not enumerated flags. I'm thinking of the best non-intrusive way to correct it.

As Ben noted, you cannot change those enumerated values, as they are already part of the ABI.

For ETHTOOL_GSSET_INFO, you initialize the sset_mask like this:

Quoted 7 lines

info.sset_mask = (1ULL << ETH_SS_NTUPLE_FILTERS);

A multiple initialization would look like this:

info.sset_mask = (1ULL << ETH_SS_NTUPLE_FILTERS) | (1ULL << ETH_SS_STATS) |
(1ULL << ETH_SS_PRIV_FLAGS);

Additionally, upon ioctl(2) completion, info.sset_mask will contain (1<<ETH_SS_NTUPLE_FILTERS) if and only if the kernel ops->get_sset_count() function call returned successfully.

Thus, the absence of that bit in info.sset_mask indicates the driver returned failure.

This condition needs to be checked in your userspace ethtool patch.

Jeff

#10 lore raw
In reply to Jeff Garzik
Quoted 48 lines · Jeff Garzik
Quoted 47 lines · Peter P Waskiewicz Jr
Quoted 46 lines · Jeff Garzik
Quoted 7 lines · Jeff Kirsher

From: Peter Waskiewicz<[email protected]>

The drvinfo struct should include the number of strings that get_rx_ntuple will return. It will be variable if an underlying driver implements its own get_rx_ntuple routine, so userspace needs to know how much data is coming.

Quoted 7 lines

---

   include/linux/ethtool.h |    1 +
   net/core/ethtool.c      |    3 +++
   2 files changed, 4 insertions(+), 0 deletions(-)

(resending reply, standard patch-sending box is having trouble sending to vger)

As noted in the other email, your patch breaks ABI. The proper path is to decrease the size of reserved struct member, and NOT shift the offset of other members.

However, perhaps consider the following patch for returning n-tuple count, for four reasons:

1) space in ethtool_drvinfo is limited

2) the patch below permits trivial string set addition, without ABI changes beyond adding a new ETH_SS_xxx constant.

3) the patch below permits direct access to ops->get_sset_count(), rather than implicit access via ethtool_drvinfo

4) ethtool_drvinfo interface does not permit indication of ops->get_sset_count() failure, versus returning zero value. The patch below does so, via output sset_mask.

WARNING: this patch is compile-tested only.

NOTE: I added a cosmetic fix to ETHTOOL_[GS]RXNTUPLE constants, making their indentation consistent with the rest of the list of constants.

Quoted 8 lines
Quoted 5 lines
Quoted 1 line

I'm updating your patch since I found an issue. The mask is passing in the ETH_SS_* flags, but then they're treated as bits, not enumerated flags. I'm thinking of the best non-intrusive way to correct it.

As Ben noted, you cannot change those enumerated values, as they are already part of the ABI.

Apologies. My outstanding IT support marked your and Ben's emails as spam, and moved it aside where I couldn't see them until this morning...

Quoted 5 lines

For ETHTOOL_GSSET_INFO, you initialize the sset_mask like this:

	info.sset_mask = (1ULL << ETH_SS_NTUPLE_FILTERS);
Quoted 8 lines

A multiple initialization would look like this:

	info.sset_mask = (1ULL << ETH_SS_NTUPLE_FILTERS) |
			(1ULL << ETH_SS_STATS) |
			(1ULL << ETH_SS_PRIV_FLAGS);

Do you still see an issue in my suggested code, now that sset_mask confusion is cleared up?

This sounds like it makes sense. I'll make the changes (and the userspace change you suggested in another mail) and get new patches out.

Quoted 4 lines

Regards,

Jeff

#11 lore raw
In reply to Jeff Kirsher
Quoted 9 lines · Jeff Kirsher

From: Peter Waskiewicz <[email protected]>

The drvinfo struct should include the number of strings that get_rx_ntuple will return. It will be variable if an underlying driver implements its own get_rx_ntuple routine, so userspace needs to know how much data is coming.

As noted in the other email, your patch breaks ABI. The proper path is to decrease the size of reserved struct member, and NOT shift the offset of other members.

However, perhaps consider the following patch for returning n-tuple count, for four reasons:

1) space in ethtool_drvinfo is limited

2) the patch below permits trivial string set addition, without ABI changes beyond adding a new ETH_SS_xxx constant.

3) the patch below permits direct access to ops->get_sset_count(), rather than implicit access via ethtool_drvinfo

4) ethtool_drvinfo interface does not permit indication of ops->get_sset_count() failure, versus returning zero value. The patch below does so, via output sset_mask.

WARNING: this patch is compile-tested only.

NOTE: I added a cosmetic fix to ETHTOOL_[GS]RXNTUPLE constants, making their indentation consistent with the rest of the list of constants.

include/linux/ethtool.h +14 -3
index f7992a2..04ecf18 100644
@@ -254,6 +254,17 @@ struct ethtool_gstrings {
 	__u8	data[0];
 };
 
+struct ethtool_sset_info {
+	__u32	cmd;		/* ETHTOOL_GSSET_INFO */
+	__u32	reserved;
+	__u64	sset_mask;	/* input: each bit selects an sset to query */
+				/* output: each bit a returned sset */
+	__u32	data[0];	/* ETH_SS_xxx count, in order, based on bits
+				   in sset_mask.  One bit implies one
+				   __u32, two bits implies two
+				   __u32's, etc. */
+};
+
 enum ethtool_test_flags {
 	ETH_TEST_FL_OFFLINE	= (1 << 0),	/* online / offline */
 	ETH_TEST_FL_FAILED	= (1 << 1),	/* test passed / failed */
@@ -607,9 +618,9 @@ struct ethtool_ops {
 #define	ETHTOOL_SRXCLSRLINS	0x00000032 /* Insert RX classification rule */
 #define	ETHTOOL_FLASHDEV	0x00000033 /* Flash firmware to device */
 #define	ETHTOOL_RESET		0x00000034 /* Reset hardware */
-
-#define ETHTOOL_SRXNTUPLE	0x00000035 /* Add an n-tuple filter to device */
-#define ETHTOOL_GRXNTUPLE	0x00000036 /* Get n-tuple filters from device */
+#define	ETHTOOL_SRXNTUPLE	0x00000035 /* Add an n-tuple filter to device */
+#define	ETHTOOL_GRXNTUPLE	0x00000036 /* Get n-tuple filters from device */
+#define	ETHTOOL_GSSET_INFO	0x00000037 /* Get string set info */
 
 /* compatibility with older code */
 #define SPARC_ETH_GSET		ETHTOOL_GSET
net/core/ethtool.c +66 -0

78 lines not on this page · raw message