Skip to contentSkip to conversation

Add lspci support for Enhanced Allocation Capability.

PATCH Superseded David Daney 4 participants 5 messages Dec 15, 2015 – Dec 23, 2015

Subsystems (2)

v3 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

PATCH v3 1/1 Add lspci support for Enhanced Allocation Capability.

From: David Daney <[email protected]>

The PCISIG recently added the Enhanced Allocation Capability. Decode it in lspci.
---
The specification is currently located here:

https://pcisig.com/sites/default/files/specification_documents/ECN_Enhanced_Allocation_23_Oct_2014_Final.pdf

Bjorn Helgaas recently merged Linux kernel support for EA, and Cavium ThunderX processors have implemented it. This patch gives us a pretty view of the capability structure.

Changes from v1: Add and use more symbolic names. Use FLAG() macro where appropiate. Read subordinate bus number from proper offset.

Changes from v2: Reword and reformat some messages.

Thanks,
David Daney

 lib/header.h |   8 +++
 ls-caps.c    | 156 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 164 insertions(+)
lib/header.h +8 -0
index f7cdee7..b8f7dc1 100644
@@ -203,6 +203,7 @@
 #define  PCI_CAP_ID_MSIX	0x11	/* MSI-X */
 #define  PCI_CAP_ID_SATA	0x12	/* Serial-ATA HBA */
 #define  PCI_CAP_ID_AF		0x13	/* Advanced features of PCI devices integrated in PCIe root cplx */
+#define  PCI_CAP_ID_EA		0x14	/* Enhanced Allocation */
 #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
@@ -906,6 +907,13 @@
 #define PCI_SATA_HBA_BARS	4
 #define PCI_SATA_HBA_REG0	8
 
+/* Enhanced Allocation (EA) */
+#define PCI_EA_CAP_TYPE1_SECONDARY	4
+#define PCI_EA_CAP_TYPE1_SUBORDINATE	5
+/* EA Entry header */
+#define PCI_EA_CAP_ENT_WRITABLE	0x40000000	/* Writable: 1 = RW, 0 = HwInit */
+#define PCI_EA_CAP_ENT_ENABLE	0x80000000	/* Enable for this entry */
+
 /*** Definitions of extended capabilities ***/
 
 /* Advanced Error Reporting */
ls-caps.c +156 -0

168 lines not on this page · raw message

Signature · 1 line
1.8.3.1
#1 lore raw
Quoted 1 line

From 484344fef7e5c96c54576586ce81e6209a48b92c Mon Sep 17 00:00:00 2001

From: "Sean O. Stalley" <[email protected]>
Date: Wed, 16 Dec 2015 10:31:51 -0800
Subject: [PATCH 1/1] Support region sizes that are not power-of-2

Enhanced Allocation allows PCI devices to reserve regions with DWORD granularity. Make sure lspci doesn't tuncate least significant bits of the region size.

ex: a 2000 byte region should display [size=2000] instead of [size=1K]

---

Hi David,

Here are some additional changes I think we should make to lspci in order to support EA.

The truncation problem was a simple fix.

Regions described by EA are being mislabeled as [virtual]. I spent some time today trying to write a patch that fixes this issue. I couldn't come up with a clean solution, so I just added a comment.

Let me know if you have any questions.

Thanks,
Sean

 lspci.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
lspci.c +2 -2
index fe7b7fe..d483f3a 100644
@@ -336,7 +336,7 @@ show_size(pciaddr_t x)
   if (!x)
     return;
   for (i = 0; i < (sizeof(suffix) / sizeof(*suffix) - 1); i++) {
-    if (x < 1024)
+    if (x % 1024)
       break;
     x /= 1024;
   }
@@ -366,7 +366,7 @@ show_bases(struct device *d, int cnt)
 	putchar('\t');
       if (pos && !flg)			/* Reported by the OS, but not by the device */
 	{
-	  printf("[virtual] ");
+	  printf("[virtual] ");		/* could also be enhanced */
 	  flg = pos;
 	  virtual = 1;
 	}
Signature · 1 line
1.9.1
#2 lore raw

Sean,

I think this is a good addition.

I don't plan on merging it into my patch, but instead think it should be applied separately

Thanks,
David Daney

Quoted 54 lines · Sean O. Stalley

From 484344fef7e5c96c54576586ce81e6209a48b92c Mon Sep 17 00:00:00 2001
From: "Sean O. Stalley" <[email protected]>
Date: Wed, 16 Dec 2015 10:31:51 -0800
Subject: [PATCH 1/1] Support region sizes that are not power-of-2

Enhanced Allocation allows PCI devices to reserve regions with DWORD granularity. Make sure lspci doesn't tuncate least significant bits of the region size.

ex: a 2000 byte region should display [size=2000] instead of [size=1K]

---

Hi David,

Here are some additional changes I think we should make to lspci in order to support EA.

The truncation problem was a simple fix.

Regions described by EA are being mislabeled as [virtual]. I spent some time today trying to write a patch that fixes this issue. I couldn't come up with a clean solution, so I just added a comment.

Let me know if you have any questions.

Thanks,
Sean

  lspci.c | 4 ++--
  1 file changed, 2 insertions(+), 2 deletions(-)
lspci.c +2 -2
index fe7b7fe..d483f3a 100644
@@ -336,7 +336,7 @@ show_size(pciaddr_t x)
    if (!x)
      return;
    for (i = 0; i < (sizeof(suffix) / sizeof(*suffix) - 1); i++) {
-    if (x < 1024)
+    if (x % 1024)
        break;
      x /= 1024;
    }
@@ -366,7 +366,7 @@ show_bases(struct device *d, int cnt)
  	putchar('\t');
        if (pos && !flg)			/* Reported by the OS, but not by the device */
  	{
-	  printf("[virtual] ");
+	  printf("[virtual] ");		/* could also be enhanced */
  	  flg = pos;
  	  virtual = 1;
  	}
#3 lore raw
In reply to David Daney

Hi!

Quoted 7 lines

From: David Daney <[email protected]>

The PCISIG recently added the Enhanced Allocation Capability. Decode it in lspci.
---
The specification is currently located here:

Quoted 5 lines

https://pcisig.com/sites/default/files/specification_documents/ECN_Enhanced_Allocation_23_Oct_2014_Final.pdf

Bjorn Helgaas recently merged Linux kernel support for EA, and Cavium ThunderX processors have implemented it. This patch gives us a pretty view of the capability structure.

I will merge this soon.

Could you please provide a hex dump of a config space with this capability? (See the test cases in pciutils/tests/.)

Have a nice fortnight

Signature · 3 lines
Martin `MJ' Mares                          <[email protected]>   http://mj.ucw.cz/
Faculty of Math and Physics, Charles University, Prague, Czech Rep., Earth
Q: How many Prolog programmers does it take to change a light bulb?  A: No.
#4 lore raw
Quoted 5 lines · Martin Mares

I will merge this soon.

Could you please provide a hex dump of a config space with this capability? (See the test cases in pciutils/tests/.)

Attached is a hex dump from a QEMU model using EA entries in leiu of BARs.

I also attached the output of lspci, so you can see how the Region fields are displayed. Note that EA memory regions get tagged as [virtual].

I attached 2 outputs to show the truncation case I was talking about [1]. lt_size is the output when (x < 1024) is used in show_size(), whereas mod_size is the output when (x % 1024) is used.

Thanks for taking a look,
Sean

[1]
diff mod_size lt_size
8c8
< Region 0: [virtual] Memory at 80000000 (32-bit, non-prefetchable) [size=131076]
---

Quoted 1 line
	Region 0: [virtual] Memory at 80000000 (32-bit, non-prefetchable) [size=128K]

00:04.0 Ethernet controller: Intel Corporation 82540EM Gigabit Ethernet Controller (rev 03)
Subsystem: Red Hat, Inc QEMU Virtual Machine
Physical Slot: 4
Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR+ FastB2B- DisINTx-
Status: Cap+ 66MHz- UDF- FastB2B- ParErr- DEVSEL=fast >TAbort- <TAbort- <MAbort- >SERR- <PERR- INTx-
Latency: 0
Interrupt: pin A routed to IRQ 11
Region 0: [virtual] Memory at 80000000 (32-bit, non-prefetchable) [size=131076] Region 1: I/O ports at 0a00 [size=68]
Capabilities: [40] Enhanced Allocation (EA): NumEntries=2
Entry 0: Enable+ Writable- EntrySize=2
BAR Equivalent Indicator: BAR 0
PrimaryProperties: memory space, non-prefetchable
SecondaryProperties: entry unavailable for use, PrimaryProperties should be used
Base: 80000000
MaxOffset: 00020003
Entry 1: Enable+ Writable- EntrySize=2
BAR Equivalent Indicator: BAR 1
PrimaryProperties: I/O space
SecondaryProperties: entry unavailable for use, PrimaryProperties should be used
Base: 00000a00
MaxOffset: 00000043
Kernel driver in use: e1000

00:04.0 Ethernet controller: Intel Corporation 82540EM Gigabit Ethernet Controller (rev 03)
Subsystem: Red Hat, Inc QEMU Virtual Machine
Physical Slot: 4
Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR+ FastB2B- DisINTx-
Status: Cap+ 66MHz- UDF- FastB2B- ParErr- DEVSEL=fast >TAbort- <TAbort- <MAbort- >SERR- <PERR- INTx-
Latency: 0
Interrupt: pin A routed to IRQ 11
Region 0: [virtual] Memory at 80000000 (32-bit, non-prefetchable) [size=128K] Region 1: I/O ports at 0a00 [size=68]
Capabilities: [40] Enhanced Allocation (EA): NumEntries=2
Entry 0: Enable+ Writable- EntrySize=2
BAR Equivalent Indicator: BAR 0
PrimaryProperties: memory space, non-prefetchable
SecondaryProperties: entry unavailable for use, PrimaryProperties should be used
Base: 80000000
MaxOffset: 00020003
Entry 1: Enable+ Writable- EntrySize=2
BAR Equivalent Indicator: BAR 1
PrimaryProperties: I/O space
SecondaryProperties: entry unavailable for use, PrimaryProperties should be used
Base: 00000a00
MaxOffset: 00000043
Kernel driver in use: e1000

00:04.0 Ethernet controller: Intel Corporation 82540EM Gigabit Ethernet Controller (rev 03) 00: 86 80 0e 10 07 01 10 00 03 00 00 02 00 00 00 00
10: 00 00 00 00 01 00 00 00 00 00 00 00 00 00 00 00
20: 00 00 00 00 00 00 00 00 00 00 00 00 f4 1a 00 11
30: 00 00 00 00 40 00 00 00 00 00 00 00 0b 01 00 00
40: 14 00 02 00 02 00 ff 80 00 00 00 80 00 00 02 00
50: 12 02 ff 80 00 0a 00 00 40 00 00 00 00 00 00 00
60: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
70: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
80: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
90: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
a0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
b0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
c0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
d0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
e0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
f0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00