Skip to contentSkip to conversation

pciutils: Support for MSI-X capability

PATCH In review Roland Dreier 3 participants 8 messages Jun 26, 2004 – Jun 27, 2004 Library code

lore
Share & tools
lore

Roland Dreier

author
#0 lore raw

PATCH 1/1 pciutils: Support for MSI-X capability

Hi, here is a patch to pciutils that adds parsing of MSI-X capability entries. With this patch, an MSI-X capability will be dumped with -v as

Capabilities: [40] MSI-X: Enable- Mask- TabSize=32

and with -vv as

	Capabilities: [40] MSI-X: Enable- Mask- TabSize=32
		Vector table: BAR=0 offset=00082000
		PBA: BAR=0 offset=00082200

Please let me know if you need any changes/fixes before you can apply.

Thanks,
Roland

lib/header.h +9 -0
Index: pciutils-2.1.99-test5/lib/header.h
===================================================================
@@ -185,6 +185,7 @@
 #define  PCI_CAP_ID_CHSWP	0x06	/* CompactPCI HotSwap */
 #define  PCI_CAP_ID_PCIX        0x07    /* PCI-X */
 #define  PCI_CAP_ID_HT          0x08    /* HyperTransport */
+#define  PCI_CAP_ID_MSIX        0x11    /* MSI-X */
 #define PCI_CAP_LIST_NEXT	1	/* Next capability in the list */
 #define PCI_CAP_FLAGS		2	/* Capability defined flags (16 bits) */
 #define PCI_CAP_SIZEOF		4

@@ -661,6 +662,14 @@
 #define PCI_HT_RM_CNT1		10	/* Retry Count 1 Register */
 #define PCI_HT_RM_SIZEOF	12
 
+/* MSI-X */
+#define  PCI_MSIX_ENABLE  0x8000
+#define  PCI_MSIX_MASK    0x4000
+#define  PCI_MSIX_TABSIZE 0x03ff
+#define PCI_MSIX_TABLE    4
+#define PCI_MSIX_PBA      8
+#define  PCI_MSIX_BIR     0x7
+
 /*
  * The PCI interface treats multi-function devices as independent
  * devices.  The slot/function address of each device is encoded
lspci.c +25 -0
Index: pciutils-2.1.99-test5/lspci.c
===================================================================
@@ -924,6 +924,28 @@
 }
 
 static void
+show_msix(struct device *d, int where, int cap)
+{
+  u32 off;
+
+  printf("MSI-X: Enable%c Mask%c TabSize=%d\n",
+	 FLAG(cap, PCI_MSIX_ENABLE),
+	 FLAG(cap, PCI_MSIX_MASK),
+	 (cap & PCI_MSIX_TABSIZE) + 1);
+  if (verbose < 2)
+    return;
+
+  config_fetch(d, where + PCI_MSIX_TABLE, 4);
+  off = get_conf_long(d, where + PCI_MSIX_TABLE);
+  printf("\t\tVector table: BAR=%d offset=%08x\n",
+	 off & PCI_MSIX_BIR, off & ~PCI_MSIX_BIR);
+  config_fetch(d, where + PCI_MSIX_PBA, 4);
+  off = get_conf_long(d, where + PCI_MSIX_PBA);
+  printf("\t\tPBA: BAR=%d offset=%08x\n",
+	 off & PCI_MSIX_BIR, off & ~PCI_MSIX_BIR);
+}
+
+static void
 show_slotid(int cap)
 {
   int esr = cap & 0xff;

@@ -982,6 +1004,9 @@
 	    case PCI_CAP_ID_HT:
 	      show_ht(d, where, cap);
 	      break;
+	    case PCI_CAP_ID_MSIX:
+	      show_msix(d, where, cap);
+	      break;
 	    default:
 	      printf("#%02x [%04x]\n", id, cap);
 	    }
#1 lore raw
Quoted 2 lines · Roland Dreier

Hi, here is a patch to pciutils that adds parsing of MSI-X capability entries. With this patch, an MSI-X capability will be dumped with -v as

Thanks for doing this, Roland. I was going to get to it in a few days. Unfortunately, it's going to conflict in a few textual ways with one or two of the [RFC] patches I posted.

Martin, how can we make this easiest for you? Do you want to merge Roland's fully-fledged MSI-X patch and put out a new -test release that I can send half-baked PCI-E patches against until everybody's happy with the outcome?

Signature · 6 lines
"Next the statesmen will invent cheap lies, putting the blame upon 
the nation that is attacked, and every man will be glad of those
conscience-soothing falsities, and will diligently study them, and refuse
to examine any refutations of them; and thus he will by and by convince 
himself that the war is just, and will thank God for the better sleep 
he enjoys after this process of grotesque self-deception." -- Mark Twain
author
#2 lore raw

Matthew> Martin, how can we make this easiest for you? Do you Matthew> want to merge Roland's fully-fledged MSI-X patch and put Matthew> out a new -test release that I can send half-baked PCI-E Matthew> patches against until everybody's happy with the outcome?

Ahh... I'll hold off on writing my PCI-e capability parser then :)

- Roland

#3 lore raw
Quoted 6 lines · Roland Dreier

Matthew> Martin, how can we make this easiest for you? Do you Matthew> want to merge Roland's fully-fledged MSI-X patch and put Matthew> out a new -test release that I can send half-baked PCI-E Matthew> patches against until everybody's happy with the outcome?

Ahh... I'll hold off on writing my PCI-e capability parser then :)

Did you not see the one I posted to linux-pci yesterday? As I say, it's only half-done. I wanted to get a feel for whether people like the direction I'm taking.

Unfortunately, I can't see a web archive of linux-pci anywhere -- even on MArc. I can forward the patches though.

Signature · 6 lines
"Next the statesmen will invent cheap lies, putting the blame upon 
the nation that is attacked, and every man will be glad of those
conscience-soothing falsities, and will diligently study them, and refuse
to examine any refutations of them; and thus he will by and by convince 
himself that the war is just, and will thank God for the better sleep 
he enjoys after this process of grotesque self-deception." -- Mark Twain
author
#4 lore raw

Matthew> Did you not see the one I posted to linux-pci yesterday? Matthew> As I say, it's only half-done. I wanted to get a feel Matthew> for whether people like the direction I'm taking.

Matthew> Unfortunately, I can't see a web archive of linux-pci Matthew> anywhere -- even on MArc. I can forward the patches Matthew> though.

Sorry, I'm not subscribed to linux-pci, so I missed them.

I would be interested in taking a look at your patches. Are you handling the full 4K extended config space, or just the PCI-e capability in the standard config space?

Thanks,
Roland

#5 lore raw
Quoted 8 lines · Roland Dreier

Matthew> Did you not see the one I posted to linux-pci yesterday? Matthew> As I say, it's only half-done. I wanted to get a feel Matthew> for whether people like the direction I'm taking.

Matthew> Unfortunately, I can't see a web archive of linux-pci Matthew> anywhere -- even on MArc. I can forward the patches Matthew> though.

Quoted 5 lines

Sorry, I'm not subscribed to linux-pci, so I missed them.

I would be interested in taking a look at your patches. Are you handling the full 4K extended config space, or just the PCI-e capability in the standard config space?

OK. I'll send them to you privately as replies to this mail.

The first patch (which needs a little tightening up) adds support for accessing and printing all 4k of configuration space with the -xxx option.

The second patch (incomplete) adds support for the Express capability. There's some stuff it doesn't decode yet. I'll be updating this patch on Monday, I expect.

The third patch (even more incomplete) adds infrastructure for walking and decoding extended capabilities.

Signature · 6 lines
"Next the statesmen will invent cheap lies, putting the blame upon 
the nation that is attacked, and every man will be glad of those
conscience-soothing falsities, and will diligently study them, and refuse
to examine any refutations of them; and thus he will by and by convince 
himself that the war is just, and will thank God for the better sleep 
he enjoys after this process of grotesque self-deception." -- Mark Twain
#6 lore raw
In reply to Roland Dreier

Hi!

Quoted 5 lines

Hi, here is a patch to pciutils that adds parsing of MSI-X capability entries. With this patch, an MSI-X capability will be dumped with -v as

Capabilities: [40] MSI-X: Enable- Mask- TabSize=32

Quoted 7 lines

and with -vv as

	Capabilities: [40] MSI-X: Enable- Mask- TabSize=32
		Vector table: BAR=0 offset=00082000
		PBA: BAR=0 offset=00082200

Please let me know if you need any changes/fixes before you can apply.

Applied and will appear in -test6 soon.

Could you please add a couple of comments to the capability defines in header.h?

Have a nice fortnight

Signature · 3 lines
Martin `MJ' Mares   <[email protected]>   http://atrey.karlin.mff.cuni.cz/~mj/
Faculty of Math and Physics, Charles University, Prague, Czech Rep., Earth
Q: Do you believe in One God? A: Yes, up to isomorphism.
author
#7 lore raw

Martin> Applied and will appear in -test6 soon.

Martin> Could you please add a couple of comments to the Martin> capability defines in header.h?

Thanks, sorry about leaving out the comments. Here is a patch. I also added _CAP_ to the capability register defines so that I match the naming convention better.

Best,
Roland

lib/header.h +6 -6
Index: pciutils-2.1.99-test6/lib/header.h
===================================================================
@@ -727,12 +727,12 @@
 #define PCI_EXP_RTSTA		0x20	/* Root Status */
 
 /* MSI-X */
-#define  PCI_MSIX_ENABLE	0x8000
-#define  PCI_MSIX_MASK		0x4000
-#define  PCI_MSIX_TABSIZE	0x03ff
-#define PCI_MSIX_TABLE		4
-#define PCI_MSIX_PBA		8
-#define  PCI_MSIX_BIR		0x7
+#define  PCI_MSIX_CAP_ENABLE	0x8000	/* MSI-X enabled */
+#define  PCI_MSIX_CAP_MASK	0x4000	/* All vectors masked */
+#define  PCI_MSIX_CAP_TABSIZE	0x03ff	/* Mask for (table_size - 1) */
+#define PCI_MSIX_TABLE		4	/* Vector table offset/BAR register */
+#define PCI_MSIX_PBA		8	/* PBA offset/BAR register */
+#define  PCI_MSIX_BIR_MASK	0x7	/* Mask of BAR index for table/PBA */
 
 /*
  * The PCI interface treats multi-function devices as independent
lspci.c +5 -5
Index: pciutils-2.1.99-test6/lspci.c
===================================================================
@@ -990,18 +990,18 @@
   u32 off;
 
   printf("MSI-X: Enable%c Mask%c TabSize=%d\n",
-	 FLAG(cap, PCI_MSIX_ENABLE),
-	 FLAG(cap, PCI_MSIX_MASK),
-	 (cap & PCI_MSIX_TABSIZE) + 1);
+	 FLAG(cap, PCI_MSIX_CAP_ENABLE),
+	 FLAG(cap, PCI_MSIX_CAP_MASK),
+	 (cap & PCI_MSIX_CAP_TABSIZE) + 1);
   if (verbose < 2 || !config_fetch(d, where + PCI_MSIX_TABLE, 8))
     return;
 
   off = get_conf_long(d, where + PCI_MSIX_TABLE);
   printf("\t\tVector table: BAR=%d offset=%08x\n",
-	 off & PCI_MSIX_BIR, off & ~PCI_MSIX_BIR);
+	 off & PCI_MSIX_BIR_MASK, off & ~PCI_MSIX_BIR_MASK);
   off = get_conf_long(d, where + PCI_MSIX_PBA);
   printf("\t\tPBA: BAR=%d offset=%08x\n",
-	 off & PCI_MSIX_BIR, off & ~PCI_MSIX_BIR);
+	 off & PCI_MSIX_BIR_MASK, off & ~PCI_MSIX_BIR_MASK);
 }
 
 static void