Skip to contentSkip to conversation

vfio_pci: Add local source directory as include

PATCH In review Laura Abbott 6 participants 16 messages Jan 4, 2019 – Jan 8, 2019

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

PATCH 1/1 vfio_pci: Add local source directory as include

Commit 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver") introduced a trace.h file in the local directory but missed adding the local include path, resulting in compilation failures with tracepoints:

In file included from drivers/vfio/pci/trace.h:102,
from drivers/vfio/pci/vfio_pci_nvlink2.c:29:
./include/trace/define_trace.h:89:42: fatal error: ./trace.h: No such file or directory
#include TRACE_INCLUDE(TRACE_INCLUDE_FILE)

Fix this by adjusting the include path.

Fixes 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver")

---
I'd still like to echo my sentiment that this should not be a def_bool. We hit this error on our internal testing and we couldn't even turn off the driver until we fixed this.

---
 drivers/vfio/pci/Makefile | 1 +
 1 file changed, 1 insertion(+)
drivers/vfio/pci/Makefile +1 -0
index 9662c063a6b1..08d4676a8495 100644
@@ -1,3 +1,4 @@
+ccflags-y                               += -I$(src)
 
 vfio-pci-y := vfio_pci.o vfio_pci_intrs.o vfio_pci_rdwr.o vfio_pci_config.o
 vfio-pci-$(CONFIG_VFIO_PCI_IGD) += vfio_pci_igd.o
Signature · 1 line
2.20.1
Reviewed-by on v1 1/1
#1 lore raw
Quoted 10 lines · Laura Abbott

Commit 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver") introduced a trace.h file in the local directory but missed adding the local include path, resulting in compilation failures with tracepoints:

In file included from drivers/vfio/pci/trace.h:102,
from drivers/vfio/pci/vfio_pci_nvlink2.c:29:
./include/trace/define_trace.h:89:42: fatal error: ./trace.h: No such file or directory
#include TRACE_INCLUDE(TRACE_INCLUDE_FILE)

Quoted 4 lines

Fix this by adjusting the include path.

Fixes 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver")
Quoted 18 lines

---
I'd still like to echo my sentiment that this should not be a def_bool. We hit this error on our internal testing and we couldn't even turn off the driver until we fixed this.

---
 drivers/vfio/pci/Makefile | 1 +
 1 file changed, 1 insertion(+)
drivers/vfio/pci/Makefile +1 -0
index 9662c063a6b1..08d4676a8495 100644
@@ -1,3 +1,4 @@
+ccflags-y                               += -I$(src)
 
 vfio-pci-y := vfio_pci.o vfio_pci_intrs.o vfio_pci_rdwr.o vfio_pci_config.o
 vfio-pci-$(CONFIG_VFIO_PCI_IGD) += vfio_pci_igd.o
Signature · 1 line
Alexey
#2 lore raw
In reply to Laura Abbott
Quoted 10 lines · Laura Abbott

Commit 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver") introduced a trace.h file in the local directory but missed adding the local include path, resulting in compilation failures with tracepoints:

In file included from drivers/vfio/pci/trace.h:102,
from drivers/vfio/pci/vfio_pci_nvlink2.c:29:
./include/trace/define_trace.h:89:42: fatal error: ./trace.h: No such file or directory
#include TRACE_INCLUDE(TRACE_INCLUDE_FILE)

Quoted 8 lines

Fix this by adjusting the include path.

Fixes 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver")

---
I'd still like to echo my sentiment that this should not be a def_bool. We hit this error on our internal testing and we couldn't even turn off the driver until we fixed this.

I assume there's some reason you can't commit a patch to your tree to change it to bool, or turn it off entirely? That would change the SHA which is perhaps reason enough.

In general we have far too many options and most of them never get turned off (or on), so it just creates testing/bug surface for not much benefit. This is one that will probably be turned on in all distro kernels for example.

But I have no real objection to making it user configurable.

Alex I assume you'll merge this fix via the vfio tree?

cheers

Quoted 11 lines
drivers/vfio/pci/Makefile +1 -0
index 9662c063a6b1..08d4676a8495 100644
@@ -1,3 +1,4 @@
+ccflags-y                               += -I$(src)
 
 vfio-pci-y := vfio_pci.o vfio_pci_intrs.o vfio_pci_rdwr.o vfio_pci_config.o
 vfio-pci-$(CONFIG_VFIO_PCI_IGD) += vfio_pci_igd.o
Signature · 1 line
2.20.1
#3 lore raw
Quoted 44 lines · Michael Ellerman
Quoted 10 lines · Laura Abbott

Commit 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver") introduced a trace.h file in the local directory but missed adding the local include path, resulting in compilation failures with tracepoints:

In file included from drivers/vfio/pci/trace.h:102,
from drivers/vfio/pci/vfio_pci_nvlink2.c:29:
./include/trace/define_trace.h:89:42: fatal error: ./trace.h: No such file or directory
#include TRACE_INCLUDE(TRACE_INCLUDE_FILE)

Quoted 8 lines

Fix this by adjusting the include path.

Fixes 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver")

---
I'd still like to echo my sentiment that this should not be a def_bool. We hit this error on our internal testing and we couldn't even turn off the driver until we fixed this.

I assume there's some reason you can't commit a patch to your tree to change it to bool, or turn it off entirely? That would change the SHA which is perhaps reason enough.

In general we have far too many options and most of them never get turned off (or on), so it just creates testing/bug surface for not much benefit. This is one that will probably be turned on in all distro kernels for example.

But I have no real objection to making it user configurable.

Alex I assume you'll merge this fix via the vfio tree?

cheers

Quoted 7 lines
drivers/vfio/pci/Makefile +1 -0
index 9662c063a6b1..08d4676a8495 100644
@@ -1,3 +1,4 @@
+ccflags-y                               += -I$(src)
 
Quoted 4 lines
Quoted 4 lines
drivers/vfio/pci/Makefile +0 -1
 vfio-pci-y := vfio_pci.o vfio_pci_intrs.o vfio_pci_rdwr.o vfio_pci_config.o
 vfio-pci-$(CONFIG_VFIO_PCI_IGD) += vfio_pci_igd.o
--

2.20.1

Hi.

If I correctly understand the usage of TRACE_INCLUDE_PATH, the correct fix should be like follows:

drivers/vfio/pci/trace.h +1 -1
index 228ccdb..4d13e51 100644
@@ -94,7 +94,7 @@ TRACE_EVENT(vfio_pci_npu2_mmap,
 #endif /* _TRACE_VFIO_PCI_H */
 
 #undef TRACE_INCLUDE_PATH
-#define TRACE_INCLUDE_PATH .
+#define TRACE_INCLUDE_PATH ../../drivers/vfio/pci
 #undef TRACE_INCLUDE_FILE
 #define TRACE_INCLUDE_FILE trace
 
Signature · 2 lines
Best Regards
Masahiro Yamada
#4 lore raw
Quoted 17 lines · On Mon, 7 Jan 2019 19:12:10 +0900 Masahiro Yamada…
Quoted 16 lines · Michael Ellerman
Quoted 14 lines · Laura Abbott

Commit 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver") introduced a trace.h file in the local directory but missed adding the local include path, resulting in compilation failures with tracepoints:

In file included from drivers/vfio/pci/trace.h:102,
from drivers/vfio/pci/vfio_pci_nvlink2.c:29:
./include/trace/define_trace.h:89:42: fatal error: ./trace.h: No such file or directory
#include TRACE_INCLUDE(TRACE_INCLUDE_FILE)

Fix this by adjusting the include path.

Fixes 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver")

(...)

Quoted 28 lines
Quoted 11 lines

Alex I assume you'll merge this fix via the vfio tree?

cheers

Quoted 7 lines
drivers/vfio/pci/Makefile +1 -0
index 9662c063a6b1..08d4676a8495 100644
@@ -1,3 +1,4 @@
+ccflags-y                               += -I$(src)
 
Quoted 4 lines
Quoted 4 lines
drivers/vfio/pci/Makefile +0 -1
 vfio-pci-y := vfio_pci.o vfio_pci_intrs.o vfio_pci_rdwr.o vfio_pci_config.o
 vfio-pci-$(CONFIG_VFIO_PCI_IGD) += vfio_pci_igd.o
--

2.20.1

Hi.

If I correctly understand the usage of TRACE_INCLUDE_PATH, the correct fix should be like follows:

drivers/vfio/pci/trace.h +0 -0
index 228ccdb..4d13e51 100644

drivers/vfio/pci/trace.h TRACE_EVENT(vfio_pci_npu2_mmap, · line 99

Quoted 7 lines
drivers/vfio/pci/trace.h +1 -1
@@ -94,7 +94,7 @@ TRACE_EVENT(vfio_pci_npu2_mmap,
 #endif /* _TRACE_VFIO_PCI_H */
 
 #undef TRACE_INCLUDE_PATH
-#define TRACE_INCLUDE_PATH .
+#define TRACE_INCLUDE_PATH ../../drivers/vfio/pci
 #undef TRACE_INCLUDE_FILE
 #define TRACE_INCLUDE_FILE trace

Going from the comments in samples/trace_events/trace-events-sample.h, I think both approaches are possible, and I see both used in various places.

Personally, I'd prefer Laura's patch, as it doesn't involve hardcoding a path.

#5 lore raw
Quoted 64 lines · Cornelia Huck
Quoted 17 lines · On Mon, 7 Jan 2019 19:12:10 +0900 Masahiro Yamada…
Quoted 16 lines · Michael Ellerman
Quoted 14 lines · Laura Abbott

Commit 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver") introduced a trace.h file in the local directory but missed adding the local include path, resulting in compilation failures with tracepoints:

In file included from drivers/vfio/pci/trace.h:102,
from drivers/vfio/pci/vfio_pci_nvlink2.c:29:
./include/trace/define_trace.h:89:42: fatal error: ./trace.h: No such file or directory
#include TRACE_INCLUDE(TRACE_INCLUDE_FILE)

Fix this by adjusting the include path.

Fixes 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver")

(...)

Quoted 28 lines
Quoted 11 lines

Alex I assume you'll merge this fix via the vfio tree?

cheers

Quoted 7 lines
drivers/vfio/pci/Makefile +1 -0
index 9662c063a6b1..08d4676a8495 100644
@@ -1,3 +1,4 @@
+ccflags-y                               += -I$(src)
 
Quoted 4 lines
Quoted 4 lines
drivers/vfio/pci/Makefile +0 -1
 vfio-pci-y := vfio_pci.o vfio_pci_intrs.o vfio_pci_rdwr.o vfio_pci_config.o
 vfio-pci-$(CONFIG_VFIO_PCI_IGD) += vfio_pci_igd.o
--

2.20.1

Hi.

If I correctly understand the usage of TRACE_INCLUDE_PATH, the correct fix should be like follows:

drivers/vfio/pci/trace.h +0 -0
index 228ccdb..4d13e51 100644

drivers/vfio/pci/trace.h TRACE_EVENT(vfio_pci_npu2_mmap, · line 99

Quoted 7 lines
drivers/vfio/pci/trace.h +1 -1
@@ -94,7 +94,7 @@ TRACE_EVENT(vfio_pci_npu2_mmap,
 #endif /* _TRACE_VFIO_PCI_H */
 
 #undef TRACE_INCLUDE_PATH
-#define TRACE_INCLUDE_PATH .
+#define TRACE_INCLUDE_PATH ../../drivers/vfio/pci
 #undef TRACE_INCLUDE_FILE
 #define TRACE_INCLUDE_FILE trace

Going from the comments in samples/trace_events/trace-events-sample.h, I think both approaches are possible, and I see both used in various places.

Quoted 2 lines

Personally, I'd prefer Laura's patch, as it doesn't involve hardcoding a path.

ccflags-y += -I$(src)
would add the header search path for all files in drivers/vfio/pci/ whereas only the drivers/vfio/pci/vfio_pci_nvlink2.c needs it.

CFLAGS_vfio_pci_nvlink2.o += -I$(src)
is a bit better.
However, it is not obvious why this extra header search path is needed until you find vfio_pci_nvlink2.c including trace.h

#define TRACE_INCLUDE_PATH ../../drivers/vfio/pci
clarifies the intention because the related code is all placed in trace.h

Quoted 1 line

From the comment in include/trace/define_trace.h

TRACE_INCLUDE_PATH is relative to include/trace/define_trace.h

Signature · 2 lines
Best Regards
Masahiro Yamada
#6 lore raw
Quoted 65 lines · On Mon, 7 Jan 2019 20:39:19 +0900 Masahiro Yamada…
Quoted 64 lines · Cornelia Huck
Quoted 17 lines · On Mon, 7 Jan 2019 19:12:10 +0900 Masahiro Yamada…
Quoted 16 lines · Michael Ellerman
Quoted 14 lines · Laura Abbott

Commit 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver") introduced a trace.h file in the local directory but missed adding the local include path, resulting in compilation failures with tracepoints:

In file included from drivers/vfio/pci/trace.h:102,
from drivers/vfio/pci/vfio_pci_nvlink2.c:29:
./include/trace/define_trace.h:89:42: fatal error: ./trace.h: No such file or directory
#include TRACE_INCLUDE(TRACE_INCLUDE_FILE)

Fix this by adjusting the include path.

Fixes 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver")

(...)

Quoted 28 lines
Quoted 11 lines

Alex I assume you'll merge this fix via the vfio tree?

cheers

Quoted 7 lines
drivers/vfio/pci/Makefile +1 -0
index 9662c063a6b1..08d4676a8495 100644
@@ -1,3 +1,4 @@
+ccflags-y                               += -I$(src)
 
Quoted 4 lines
Quoted 4 lines
drivers/vfio/pci/Makefile +0 -1
 vfio-pci-y := vfio_pci.o vfio_pci_intrs.o vfio_pci_rdwr.o vfio_pci_config.o
 vfio-pci-$(CONFIG_VFIO_PCI_IGD) += vfio_pci_igd.o
--

2.20.1

Hi.

If I correctly understand the usage of TRACE_INCLUDE_PATH, the correct fix should be like follows:

drivers/vfio/pci/trace.h +0 -0
index 228ccdb..4d13e51 100644

drivers/vfio/pci/trace.h TRACE_EVENT(vfio_pci_npu2_mmap, · line 99

Quoted 7 lines
drivers/vfio/pci/trace.h +1 -1
@@ -94,7 +94,7 @@ TRACE_EVENT(vfio_pci_npu2_mmap,
 #endif /* _TRACE_VFIO_PCI_H */
 
 #undef TRACE_INCLUDE_PATH
-#define TRACE_INCLUDE_PATH .
+#define TRACE_INCLUDE_PATH ../../drivers/vfio/pci
 #undef TRACE_INCLUDE_FILE
 #define TRACE_INCLUDE_FILE trace  

Going from the comments in samples/trace_events/trace-events-sample.h, I think both approaches are possible, and I see both used in various places.

Quoted 2 lines
Quoted 2 lines

Personally, I'd prefer Laura's patch, as it doesn't involve hardcoding a path.

Numbering options for clarity:

1)

Quoted 4 lines

ccflags-y += -I$(src)
would add the header search path for all files in drivers/vfio/pci/ whereas only the drivers/vfio/pci/vfio_pci_nvlink2.c needs it.

2)

Quoted 5 lines

CFLAGS_vfio_pci_nvlink2.o += -I$(src)
is a bit better.
However, it is not obvious why this extra header search path is needed until you find vfio_pci_nvlink2.c including trace.h

3)

Quoted 7 lines

#define TRACE_INCLUDE_PATH ../../drivers/vfio/pci
clarifies the intention because the related code is all placed in trace.h

From the comment in include/trace/define_trace.h TRACE_INCLUDE_PATH is relative to include/trace/define_trace.h

In my scan of the tree, the most common solution seems to be 2) as this is essentially recommended in the sample file. 3) is well represented, with much fewer examples of 1), though it might depend how liberally we grep out or examine the use cases. Choice 1) also seems to be the most shotgun approach, adding to the search path for all files. From a maintenance perspective I agree that 2) seems more error prone, especially when the build system only catches the error on in-tree builds, something I rarely do. Therefore I'm leaning towards option 3). The hardcoded path here doesn't seem much of an issue relative to the negatives of the other approaches (how often do we move these files?) and it keeps the trace support relatively self-contained. Are there further arguments for or against these options? Otherwise who wants to formally post the TRACE_INCLUDE_PATH version? Thanks,

Alex

#7 lore raw
Quoted 92 lines · Alex Williamson
Quoted 65 lines · On Mon, 7 Jan 2019 20:39:19 +0900 Masahiro Yamada…
Quoted 64 lines · Cornelia Huck
Quoted 17 lines · On Mon, 7 Jan 2019 19:12:10 +0900 Masahiro Yamada…
Quoted 16 lines · Michael Ellerman
Quoted 14 lines · Laura Abbott

Commit 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver") introduced a trace.h file in the local directory but missed adding the local include path, resulting in compilation failures with tracepoints:

In file included from drivers/vfio/pci/trace.h:102,
from drivers/vfio/pci/vfio_pci_nvlink2.c:29:
./include/trace/define_trace.h:89:42: fatal error: ./trace.h: No such file or directory
#include TRACE_INCLUDE(TRACE_INCLUDE_FILE)

Fix this by adjusting the include path.

Fixes 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver")

(...)

Quoted 28 lines
Quoted 11 lines

Alex I assume you'll merge this fix via the vfio tree?

cheers

Quoted 7 lines
drivers/vfio/pci/Makefile +1 -0
index 9662c063a6b1..08d4676a8495 100644
@@ -1,3 +1,4 @@
+ccflags-y                               += -I$(src)
 
Quoted 4 lines
Quoted 4 lines
drivers/vfio/pci/Makefile +0 -1
 vfio-pci-y := vfio_pci.o vfio_pci_intrs.o vfio_pci_rdwr.o vfio_pci_config.o
 vfio-pci-$(CONFIG_VFIO_PCI_IGD) += vfio_pci_igd.o
--

2.20.1

Hi.

If I correctly understand the usage of TRACE_INCLUDE_PATH, the correct fix should be like follows:

drivers/vfio/pci/trace.h +0 -0
index 228ccdb..4d13e51 100644

drivers/vfio/pci/trace.h TRACE_EVENT(vfio_pci_npu2_mmap, · line 99

Quoted 7 lines
drivers/vfio/pci/trace.h +1 -1
@@ -94,7 +94,7 @@ TRACE_EVENT(vfio_pci_npu2_mmap,
 #endif /* _TRACE_VFIO_PCI_H */
 
 #undef TRACE_INCLUDE_PATH
-#define TRACE_INCLUDE_PATH .
+#define TRACE_INCLUDE_PATH ../../drivers/vfio/pci
 #undef TRACE_INCLUDE_FILE
 #define TRACE_INCLUDE_FILE trace  

Going from the comments in samples/trace_events/trace-events-sample.h, I think both approaches are possible, and I see both used in various places.

Quoted 2 lines
Quoted 2 lines

Personally, I'd prefer Laura's patch, as it doesn't involve hardcoding a path.

Numbering options for clarity:

1)

Quoted 4 lines

ccflags-y += -I$(src)
would add the header search path for all files in drivers/vfio/pci/ whereas only the drivers/vfio/pci/vfio_pci_nvlink2.c needs it.

2)

Quoted 5 lines

CFLAGS_vfio_pci_nvlink2.o += -I$(src)
is a bit better.
However, it is not obvious why this extra header search path is needed until you find vfio_pci_nvlink2.c including trace.h

3)

Quoted 5 lines

#define TRACE_INCLUDE_PATH ../../drivers/vfio/pci
clarifies the intention because the related code is all placed in trace.h

Quoted 8 lines
Quoted 2 lines

From the comment in include/trace/define_trace.h TRACE_INCLUDE_PATH is relative to include/trace/define_trace.h

In my scan of the tree, the most common solution seems to be 2) as this is essentially recommended in the sample file. 3) is well represented, with much fewer examples of 1), though it might depend how liberally we grep out or examine the use cases. Choice 1) also seems to be the most shotgun approach, adding to the search path for all files.

The shotgun approach is always used when compiling out of tree which is what many people do anyway without realizing that there are additional -I. Why do not we make in-tree behave the same way? Thanks,

Quoted 9 lines

From a
maintenance perspective I agree that 2) seems more error prone, especially when the build system only catches the error on in-tree builds, something I rarely do. Therefore I'm leaning towards option 3). The hardcoded path here doesn't seem much of an issue relative to the negatives of the other approaches (how often do we move these files?) and it keeps the trace support relatively self-contained. Are there further arguments for or against these options? Otherwise who wants to formally post the TRACE_INCLUDE_PATH version? Thanks,

Signature · 1 line
Alexey
#8 lore raw
Quoted 103 lines · On Tue, 8 Jan 2019 10:52:43 +1100 Alexey Kardashevskiy <[email protected]> wrote:
Quoted 92 lines · Alex Williamson
Quoted 65 lines · On Mon, 7 Jan 2019 20:39:19 +0900 Masahiro Yamada…
Quoted 64 lines · Cornelia Huck
Quoted 17 lines · On Mon, 7 Jan 2019 19:12:10 +0900 Masahiro Yamada…
Quoted 16 lines · Michael Ellerman
Quoted 14 lines · Laura Abbott

Commit 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver") introduced a trace.h file in the local directory but missed adding the local include path, resulting in compilation failures with tracepoints:

In file included from drivers/vfio/pci/trace.h:102,
from drivers/vfio/pci/vfio_pci_nvlink2.c:29:
./include/trace/define_trace.h:89:42: fatal error: ./trace.h: No such file or directory
#include TRACE_INCLUDE(TRACE_INCLUDE_FILE)

Fix this by adjusting the include path.

Fixes 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver")

(...)

Quoted 28 lines
Quoted 11 lines

Alex I assume you'll merge this fix via the vfio tree?

cheers

Quoted 7 lines
drivers/vfio/pci/Makefile +1 -0
index 9662c063a6b1..08d4676a8495 100644
@@ -1,3 +1,4 @@
+ccflags-y                               += -I$(src)
 
Quoted 4 lines
Quoted 4 lines
drivers/vfio/pci/Makefile +0 -1
 vfio-pci-y := vfio_pci.o vfio_pci_intrs.o vfio_pci_rdwr.o vfio_pci_config.o
 vfio-pci-$(CONFIG_VFIO_PCI_IGD) += vfio_pci_igd.o
--

2.20.1

Hi.

If I correctly understand the usage of TRACE_INCLUDE_PATH, the correct fix should be like follows:

drivers/vfio/pci/trace.h +0 -0
index 228ccdb..4d13e51 100644

drivers/vfio/pci/trace.h TRACE_EVENT(vfio_pci_npu2_mmap, · line 99

Quoted 7 lines
drivers/vfio/pci/trace.h +1 -1
@@ -94,7 +94,7 @@ TRACE_EVENT(vfio_pci_npu2_mmap,
 #endif /* _TRACE_VFIO_PCI_H */
 
 #undef TRACE_INCLUDE_PATH
-#define TRACE_INCLUDE_PATH .
+#define TRACE_INCLUDE_PATH ../../drivers/vfio/pci
 #undef TRACE_INCLUDE_FILE
 #define TRACE_INCLUDE_FILE trace    

Going from the comments in samples/trace_events/trace-events-sample.h, I think both approaches are possible, and I see both used in various places.

Quoted 2 lines
Quoted 2 lines

Personally, I'd prefer Laura's patch, as it doesn't involve hardcoding a path.

Numbering options for clarity:

1)

Quoted 4 lines

ccflags-y += -I$(src)
would add the header search path for all files in drivers/vfio/pci/ whereas only the drivers/vfio/pci/vfio_pci_nvlink2.c needs it.

2)

Quoted 5 lines

CFLAGS_vfio_pci_nvlink2.o += -I$(src)
is a bit better.
However, it is not obvious why this extra header search path is needed until you find vfio_pci_nvlink2.c including trace.h

3)

Quoted 5 lines

#define TRACE_INCLUDE_PATH ../../drivers/vfio/pci
clarifies the intention because the related code is all placed in trace.h

Quoted 8 lines
Quoted 2 lines

From the comment in include/trace/define_trace.h TRACE_INCLUDE_PATH is relative to include/trace/define_trace.h

In my scan of the tree, the most common solution seems to be 2) as this is essentially recommended in the sample file. 3) is well represented, with much fewer examples of 1), though it might depend how liberally we grep out or examine the use cases. Choice 1) also seems to be the most shotgun approach, adding to the search path for all files.

Quoted 3 lines

The shotgun approach is always used when compiling out of tree which is what many people do anyway without realizing that there are additional -I. Why do not we make in-tree behave the same way? Thanks,

Are you suggesting bandaging this individual leaf directory, as in choice 1), because it's no worse than what happens for an out-of-tree build anyway, or are you suggesting to fix the in-tree build behavior to be as inefficient as the out-of-tree behavior in general? It appears to me that options 2) and 3) are the intended solutions for this issue while 1) is more of a workaround. Thanks,

Alex

Quoted 10 lines
Quoted 9 lines

From a
maintenance perspective I agree that 2) seems more error prone, especially when the build system only catches the error on in-tree builds, something I rarely do. Therefore I'm leaning towards option 3). The hardcoded path here doesn't seem much of an issue relative to the negatives of the other approaches (how often do we move these files?) and it keeps the trace support relatively self-contained. Are there further arguments for or against these options? Otherwise who wants to formally post the TRACE_INCLUDE_PATH version? Thanks,

#9 lore raw
Quoted 106 lines · Alex Williamson
Quoted 103 lines · On Tue, 8 Jan 2019 10:52:43 +1100 Alexey Kardashevskiy <[email protected]> wrote:
Quoted 100 lines · Alex Williamson
Quoted 65 lines · On Mon, 7 Jan 2019 20:39:19 +0900 Masahiro Yamada…
Quoted 64 lines · Cornelia Huck
Quoted 17 lines · On Mon, 7 Jan 2019 19:12:10 +0900 Masahiro Yamada…
Quoted 16 lines · Michael Ellerman
Quoted 14 lines · Laura Abbott

Commit 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver") introduced a trace.h file in the local directory but missed adding the local include path, resulting in compilation failures with tracepoints:

In file included from drivers/vfio/pci/trace.h:102,
from drivers/vfio/pci/vfio_pci_nvlink2.c:29:
./include/trace/define_trace.h:89:42: fatal error: ./trace.h: No such file or directory
#include TRACE_INCLUDE(TRACE_INCLUDE_FILE)

Fix this by adjusting the include path.

Fixes 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver")

(...)

Quoted 28 lines
Quoted 11 lines

Alex I assume you'll merge this fix via the vfio tree?

cheers

Quoted 7 lines
drivers/vfio/pci/Makefile +1 -0
index 9662c063a6b1..08d4676a8495 100644
@@ -1,3 +1,4 @@
+ccflags-y                               += -I$(src)
 
Quoted 4 lines
Quoted 4 lines
drivers/vfio/pci/Makefile +0 -1
 vfio-pci-y := vfio_pci.o vfio_pci_intrs.o vfio_pci_rdwr.o vfio_pci_config.o
 vfio-pci-$(CONFIG_VFIO_PCI_IGD) += vfio_pci_igd.o
--

2.20.1

Hi.

If I correctly understand the usage of TRACE_INCLUDE_PATH, the correct fix should be like follows:

drivers/vfio/pci/trace.h +0 -0
index 228ccdb..4d13e51 100644

drivers/vfio/pci/trace.h TRACE_EVENT(vfio_pci_npu2_mmap, · line 99

Quoted 7 lines
drivers/vfio/pci/trace.h +1 -1
@@ -94,7 +94,7 @@ TRACE_EVENT(vfio_pci_npu2_mmap,
 #endif /* _TRACE_VFIO_PCI_H */
 
 #undef TRACE_INCLUDE_PATH
-#define TRACE_INCLUDE_PATH .
+#define TRACE_INCLUDE_PATH ../../drivers/vfio/pci
 #undef TRACE_INCLUDE_FILE
 #define TRACE_INCLUDE_FILE trace    

Going from the comments in samples/trace_events/trace-events-sample.h, I think both approaches are possible, and I see both used in various places.

Quoted 2 lines
Quoted 2 lines

Personally, I'd prefer Laura's patch, as it doesn't involve hardcoding a path.

Numbering options for clarity:

1)

Quoted 4 lines

ccflags-y += -I$(src)
would add the header search path for all files in drivers/vfio/pci/ whereas only the drivers/vfio/pci/vfio_pci_nvlink2.c needs it.

2)

Quoted 5 lines

CFLAGS_vfio_pci_nvlink2.o += -I$(src)
is a bit better.
However, it is not obvious why this extra header search path is needed until you find vfio_pci_nvlink2.c including trace.h

3)

Quoted 7 lines

#define TRACE_INCLUDE_PATH ../../drivers/vfio/pci
clarifies the intention because the related code is all placed in trace.h

From the comment in include/trace/define_trace.h TRACE_INCLUDE_PATH is relative to include/trace/define_trace.h

In my scan of the tree, the most common solution seems to be 2) as this is essentially recommended in the sample file. 3) is well represented, with much fewer examples of 1), though it might depend how liberally we grep out or examine the use cases. Choice 1) also seems to be the most shotgun approach, adding to the search path for all files.

Quoted 7 lines
Quoted 3 lines

The shotgun approach is always used when compiling out of tree which is what many people do anyway without realizing that there are additional -I. Why do not we make in-tree behave the same way? Thanks,

Are you suggesting bandaging this individual leaf directory, as in choice 1), because it's no worse than what happens for an out-of-tree build anyway,

I suggest making in-tree and out-of-tree behavior equal - both either fail or compile. Just makes sense to me.

Since out-of-tree adds extra -I, then there should have been a reason for that at the time (before 2005 though). Unfortunately git blame does not say why it was done this way for out-of-tree so imho the safest thing is to add -I for in-tree as well. Or ditch extra -I and do 2) or
3) - this is fine too and can count as an impressive compile optimization, although... look below :)

Quoted 2 lines

or are you suggesting to fix the in-tree build behavior to be as inefficient as the out-of-tree behavior in general?

Inefficient you say. Hm.

I tried this:

Makefile.lib renamed from Makefile.lib.old +0 -0
@@ -140,11 +140,6 @@
 # If building the kernel in a separate objtree expand all occurrences
 # of -Idir to -I$(srctree)/dir except for absolute paths (starting with

'/').

-ifeq ($(KBUILD_SRC),)
-__c_flags = $(_c_flags)
-__a_flags = $(_a_flags)
-__cpp_flags = $(_cpp_flags)
-else

# -I$(obj) locates generated .h files
# $(call addtree,-I$(obj)) locates .h files in srctree, from generated
.c files
@@ -154,7 +149,6 @@
$(call flags,_c_flags)
__a_flags = $(call flags,_a_flags)
__cpp_flags = $(call flags,_cpp_flags)
-endif

Compiled 3 times with the patch and without, "make clean ; time make -j200 vmlinux modules".

No patch:

real    4m33.047s       user    481m48.322s     sys     10m15.639s
real    4m29.038s       user    480m22.873s     sys     10m11.394s
real    4m34.373s       user    483m7.570s      sys     10m10.559s

With the patch:

real    4m32.008s       user    479m54.207s     sys     10m13.075s
real    4m30.027s       user    479m46.272s     sys     10m15.886s
real    4m31.548s       user    480m2.897s      sys     10m10.024

which is slightly faster but I guess within accuracy.

Quoted 15 lines

It appears
to me that options 2) and 3) are the intended solutions for this issue while 1) is more of a workaround. Thanks,

Alex

Quoted 9 lines
Quoted 9 lines

From a
maintenance perspective I agree that 2) seems more error prone, especially when the build system only catches the error on in-tree builds, something I rarely do. Therefore I'm leaning towards option 3). The hardcoded path here doesn't seem much of an issue relative to the negatives of the other approaches (how often do we move these files?) and it keeps the trace support relatively self-contained. Are there further arguments for or against these options? Otherwise who wants to formally post the TRACE_INCLUDE_PATH version? Thanks,

Signature · 1 line
Alexey
#10 lore raw
Quoted 122 lines · Alexey Kardashevskiy
Quoted 106 lines · Alex Williamson
Quoted 103 lines · On Tue, 8 Jan 2019 10:52:43 +1100 Alexey Kardashevskiy <[email protected]> wrote:
Quoted 100 lines · Alex Williamson
Quoted 65 lines · On Mon, 7 Jan 2019 20:39:19 +0900 Masahiro Yamada…
Quoted 64 lines · Cornelia Huck
Quoted 17 lines · On Mon, 7 Jan 2019 19:12:10 +0900 Masahiro Yamada…
Quoted 16 lines · Michael Ellerman
Quoted 14 lines · Laura Abbott

Commit 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver") introduced a trace.h file in the local directory but missed adding the local include path, resulting in compilation failures with tracepoints:

In file included from drivers/vfio/pci/trace.h:102,
from drivers/vfio/pci/vfio_pci_nvlink2.c:29:
./include/trace/define_trace.h:89:42: fatal error: ./trace.h: No such file or directory
#include TRACE_INCLUDE(TRACE_INCLUDE_FILE)

Fix this by adjusting the include path.

Fixes 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver")

(...)

Quoted 28 lines
Quoted 11 lines

Alex I assume you'll merge this fix via the vfio tree?

cheers

Quoted 7 lines
drivers/vfio/pci/Makefile +1 -0
index 9662c063a6b1..08d4676a8495 100644
@@ -1,3 +1,4 @@
+ccflags-y                               += -I$(src)
 
Quoted 4 lines
Quoted 4 lines
drivers/vfio/pci/Makefile +0 -1
 vfio-pci-y := vfio_pci.o vfio_pci_intrs.o vfio_pci_rdwr.o vfio_pci_config.o
 vfio-pci-$(CONFIG_VFIO_PCI_IGD) += vfio_pci_igd.o
--

2.20.1

Hi.

If I correctly understand the usage of TRACE_INCLUDE_PATH, the correct fix should be like follows:

drivers/vfio/pci/trace.h +0 -0
index 228ccdb..4d13e51 100644

drivers/vfio/pci/trace.h TRACE_EVENT(vfio_pci_npu2_mmap, · line 99

Quoted 7 lines
drivers/vfio/pci/trace.h +1 -1
@@ -94,7 +94,7 @@ TRACE_EVENT(vfio_pci_npu2_mmap,
 #endif /* _TRACE_VFIO_PCI_H */
 
 #undef TRACE_INCLUDE_PATH
-#define TRACE_INCLUDE_PATH .
+#define TRACE_INCLUDE_PATH ../../drivers/vfio/pci
 #undef TRACE_INCLUDE_FILE
 #define TRACE_INCLUDE_FILE trace

Going from the comments in samples/trace_events/trace-events-sample.h, I think both approaches are possible, and I see both used in various places.

Quoted 2 lines
Quoted 2 lines

Personally, I'd prefer Laura's patch, as it doesn't involve hardcoding a path.

Numbering options for clarity:

1)

Quoted 4 lines

ccflags-y += -I$(src)
would add the header search path for all files in drivers/vfio/pci/ whereas only the drivers/vfio/pci/vfio_pci_nvlink2.c needs it.

2)

Quoted 5 lines

CFLAGS_vfio_pci_nvlink2.o += -I$(src)
is a bit better.
However, it is not obvious why this extra header search path is needed until you find vfio_pci_nvlink2.c including trace.h

3)

Quoted 7 lines

#define TRACE_INCLUDE_PATH ../../drivers/vfio/pci
clarifies the intention because the related code is all placed in trace.h

From the comment in include/trace/define_trace.h TRACE_INCLUDE_PATH is relative to include/trace/define_trace.h

In my scan of the tree, the most common solution seems to be 2) as this is essentially recommended in the sample file. 3) is well represented, with much fewer examples of 1), though it might depend how liberally we grep out or examine the use cases. Choice 1) also seems to be the most shotgun approach, adding to the search path for all files.

Quoted 7 lines
Quoted 3 lines

The shotgun approach is always used when compiling out of tree which is what many people do anyway without realizing that there are additional -I. Why do not we make in-tree behave the same way? Thanks,

Are you suggesting bandaging this individual leaf directory, as in choice 1), because it's no worse than what happens for an out-of-tree build anyway,

I suggest making in-tree and out-of-tree behavior equal - both either fail or compile. Just makes sense to me.

Quoted 6 lines

Since out-of-tree adds extra -I, then there should have been a reason for that at the time (before 2005 though). Unfortunately git blame does not say why it was done this way for out-of-tree so imho the safest thing is to add -I for in-tree as well. Or ditch extra -I and do 2) or
3) - this is fine too and can count as an impressive compile optimization, although... look below :)

For the out-of-tree building,
scripts/Makefile.lib adds -I$(srctree)/$(src) and -I$(obj).

In my understanding, -I$(srctree)/$(src) is for including check-in headers from generated C files.

-I$(obj) is for including generated headers from check-in C files.

One example of the correct usage of this is,

crypto/rsa_helper.c (check-in file)

includes

crypto/rsapubkey.asn1.h (generated fle)

Those extra header search paths are unneeded for in-tree building since generated files and check-in files exist in the same tree.

You just happened to find it works for out-of-tree building.

We are talking about how to make include/trace/define_trace.h include drivers/vfio/pci/trace.h

The build system cannot (should not)
cater to such a particular case.

Quoted 72 lines
Quoted 3 lines

or are you suggesting to fix the in-tree build behavior to be as inefficient as the out-of-tree behavior in general?

Inefficient you say. Hm.

I tried this:

Makefile.lib 2019-01-08 13:09:54.199054981 +1100 renamed from Makefile.lib.old 2019-01-08 11:39:51.830983393 +1100 +0 -0
@@ -140,11 +140,6 @@
 # If building the kernel in a separate objtree expand all occurrences
 # of -Idir to -I$(srctree)/dir except for absolute paths (starting with

'/').

Makefile.lib 2019-01-08 13:09:54.199054981 +1100 renamed from Makefile.lib.old 2019-01-08 11:39:51.830983393 +1100 +0 -5
-ifeq ($(KBUILD_SRC),)
-__c_flags      = $(_c_flags)
-__a_flags      = $(_a_flags)
-__cpp_flags     = $(_cpp_flags)
-else
 
 # -I$(obj) locates generated .h files
 # $(call addtree,-I$(obj)) locates .h files in srctree, from generated

.c files

Makefile.lib 2019-01-08 13:09:54.199054981 +1100 renamed from Makefile.lib.old 2019-01-08 11:39:51.830983393 +1100 +0 -1
@@ -154,7 +149,6 @@
                  $(call flags,_c_flags)
 __a_flags      = $(call flags,_a_flags)
 __cpp_flags     = $(call flags,_cpp_flags)
-endif
 
 

Compiled 3 times with the patch and without, "make clean ; time make -j200 vmlinux modules".

No patch:

real    4m33.047s       user    481m48.322s     sys     10m15.639s
real    4m29.038s       user    480m22.873s     sys     10m11.394s
real    4m34.373s       user    483m7.570s      sys     10m10.559s

With the patch:

real    4m32.008s       user    479m54.207s     sys     10m13.075s
real    4m30.027s       user    479m46.272s     sys     10m15.886s
real    4m31.548s       user    480m2.897s      sys     10m10.024

which is slightly faster but I guess within accuracy.

Quoted 4 lines

It appears
to me that options 2) and 3) are the intended solutions for this issue while 1) is more of a workaround. Thanks,

Quoted 11 lines

Alex

Quoted 9 lines
Quoted 9 lines

From a
maintenance perspective I agree that 2) seems more error prone, especially when the build system only catches the error on in-tree builds, something I rarely do. Therefore I'm leaning towards option 3). The hardcoded path here doesn't seem much of an issue relative to the negatives of the other approaches (how often do we move these files?) and it keeps the trace support relatively self-contained. Are there further arguments for or against these options? Otherwise who wants to formally post the TRACE_INCLUDE_PATH version? Thanks,

--
Alexey

Signature · 2 lines
Best Regards
Masahiro Yamada
#11 lore raw
Quoted 156 lines · Masahiro Yamada
Quoted 122 lines · Alexey Kardashevskiy
Quoted 106 lines · Alex Williamson
Quoted 103 lines · On Tue, 8 Jan 2019 10:52:43 +1100 Alexey Kardashevskiy <[email protected]> wrote:
Quoted 100 lines · Alex Williamson
Quoted 65 lines · On Mon, 7 Jan 2019 20:39:19 +0900 Masahiro Yamada…
Quoted 64 lines · Cornelia Huck
Quoted 17 lines · On Mon, 7 Jan 2019 19:12:10 +0900 Masahiro Yamada…
Quoted 16 lines · Michael Ellerman
Quoted 14 lines · Laura Abbott

Commit 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver") introduced a trace.h file in the local directory but missed adding the local include path, resulting in compilation failures with tracepoints:

In file included from drivers/vfio/pci/trace.h:102,
from drivers/vfio/pci/vfio_pci_nvlink2.c:29:
./include/trace/define_trace.h:89:42: fatal error: ./trace.h: No such file or directory
#include TRACE_INCLUDE(TRACE_INCLUDE_FILE)

Fix this by adjusting the include path.

Fixes 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver")

(...)

Quoted 28 lines
Quoted 11 lines

Alex I assume you'll merge this fix via the vfio tree?

cheers

Quoted 7 lines
drivers/vfio/pci/Makefile +1 -0
index 9662c063a6b1..08d4676a8495 100644
@@ -1,3 +1,4 @@
+ccflags-y                               += -I$(src)
 
Quoted 4 lines
Quoted 4 lines
drivers/vfio/pci/Makefile +0 -1
 vfio-pci-y := vfio_pci.o vfio_pci_intrs.o vfio_pci_rdwr.o vfio_pci_config.o
 vfio-pci-$(CONFIG_VFIO_PCI_IGD) += vfio_pci_igd.o
--

2.20.1

Hi.

If I correctly understand the usage of TRACE_INCLUDE_PATH, the correct fix should be like follows:

drivers/vfio/pci/trace.h +0 -0
index 228ccdb..4d13e51 100644

drivers/vfio/pci/trace.h TRACE_EVENT(vfio_pci_npu2_mmap, · line 99

Quoted 7 lines
drivers/vfio/pci/trace.h +1 -1
@@ -94,7 +94,7 @@ TRACE_EVENT(vfio_pci_npu2_mmap,
 #endif /* _TRACE_VFIO_PCI_H */
 
 #undef TRACE_INCLUDE_PATH
-#define TRACE_INCLUDE_PATH .
+#define TRACE_INCLUDE_PATH ../../drivers/vfio/pci
 #undef TRACE_INCLUDE_FILE
 #define TRACE_INCLUDE_FILE trace

Going from the comments in samples/trace_events/trace-events-sample.h, I think both approaches are possible, and I see both used in various places.

Quoted 2 lines
Quoted 2 lines

Personally, I'd prefer Laura's patch, as it doesn't involve hardcoding a path.

Numbering options for clarity:

1)

Quoted 4 lines

ccflags-y += -I$(src)
would add the header search path for all files in drivers/vfio/pci/ whereas only the drivers/vfio/pci/vfio_pci_nvlink2.c needs it.

2)

Quoted 5 lines

CFLAGS_vfio_pci_nvlink2.o += -I$(src)
is a bit better.
However, it is not obvious why this extra header search path is needed until you find vfio_pci_nvlink2.c including trace.h

3)

Quoted 7 lines

#define TRACE_INCLUDE_PATH ../../drivers/vfio/pci
clarifies the intention because the related code is all placed in trace.h

From the comment in include/trace/define_trace.h TRACE_INCLUDE_PATH is relative to include/trace/define_trace.h

In my scan of the tree, the most common solution seems to be 2) as this is essentially recommended in the sample file. 3) is well represented, with much fewer examples of 1), though it might depend how liberally we grep out or examine the use cases. Choice 1) also seems to be the most shotgun approach, adding to the search path for all files.

Quoted 7 lines
Quoted 3 lines

The shotgun approach is always used when compiling out of tree which is what many people do anyway without realizing that there are additional -I. Why do not we make in-tree behave the same way? Thanks,

Are you suggesting bandaging this individual leaf directory, as in choice 1), because it's no worse than what happens for an out-of-tree build anyway,

I suggest making in-tree and out-of-tree behavior equal - both either fail or compile. Just makes sense to me.

Quoted 6 lines

Since out-of-tree adds extra -I, then there should have been a reason for that at the time (before 2005 though). Unfortunately git blame does not say why it was done this way for out-of-tree so imho the safest thing is to add -I for in-tree as well. Or ditch extra -I and do 2) or
3) - this is fine too and can count as an impressive compile optimization, although... look below :)

For the out-of-tree building,
scripts/Makefile.lib adds -I$(srctree)/$(src) and -I$(obj).

In my understanding, -I$(srctree)/$(src) is for including check-in headers from generated C files.

-I$(obj) is for including generated headers from check-in C files.

One example of the correct usage of this is,

crypto/rsa_helper.c (check-in file)

includes

crypto/rsapubkey.asn1.h (generated fle)

Those extra header search paths are unneeded for in-tree building since generated files and check-in files exist in the same tree.

Quoted 8 lines

You just happened to find it works for out-of-tree building.

We are talking about how to make include/trace/define_trace.h include drivers/vfio/pci/trace.h

The build system cannot (should not)
cater to such a particular case.

Oh well, thanks for the explanation.

Alex, how do we proceed now? I like 2) + comment as in
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/misc/ocxl/Makefile?h=v4.20#n8
better than 3) as relative paths make it dependable on what file actually includes it and it is not clear what is that and what happens if that file moves deeper in the tree (unlikely to happen though).

Signature · 1 line
Alexey
#12 lore raw
Quoted 166 lines · Alexey Kardashevskiy
Quoted 162 lines · Masahiro Yamada
Quoted 122 lines · Alexey Kardashevskiy
Quoted 106 lines · Alex Williamson
Quoted 103 lines · On Tue, 8 Jan 2019 10:52:43 +1100 Alexey Kardashevskiy <[email protected]> wrote:
Quoted 100 lines · Alex Williamson
Quoted 65 lines · On Mon, 7 Jan 2019 20:39:19 +0900 Masahiro Yamada…
Quoted 64 lines · Cornelia Huck
Quoted 17 lines · On Mon, 7 Jan 2019 19:12:10 +0900 Masahiro Yamada…
Quoted 16 lines · Michael Ellerman
Quoted 14 lines · Laura Abbott

Commit 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver") introduced a trace.h file in the local directory but missed adding the local include path, resulting in compilation failures with tracepoints:

In file included from drivers/vfio/pci/trace.h:102,
from drivers/vfio/pci/vfio_pci_nvlink2.c:29:
./include/trace/define_trace.h:89:42: fatal error: ./trace.h: No such file or directory
#include TRACE_INCLUDE(TRACE_INCLUDE_FILE)

Fix this by adjusting the include path.

Fixes 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver")

(...)

Quoted 28 lines
Quoted 11 lines

Alex I assume you'll merge this fix via the vfio tree?

cheers

Quoted 7 lines
drivers/vfio/pci/Makefile +1 -0
index 9662c063a6b1..08d4676a8495 100644
@@ -1,3 +1,4 @@
+ccflags-y                               += -I$(src)
 
Quoted 4 lines
Quoted 4 lines
drivers/vfio/pci/Makefile +0 -1
 vfio-pci-y := vfio_pci.o vfio_pci_intrs.o vfio_pci_rdwr.o vfio_pci_config.o
 vfio-pci-$(CONFIG_VFIO_PCI_IGD) += vfio_pci_igd.o
--

2.20.1

Hi.

If I correctly understand the usage of TRACE_INCLUDE_PATH, the correct fix should be like follows:

drivers/vfio/pci/trace.h +0 -0
index 228ccdb..4d13e51 100644

drivers/vfio/pci/trace.h TRACE_EVENT(vfio_pci_npu2_mmap, · line 99

Quoted 7 lines
drivers/vfio/pci/trace.h +1 -1
@@ -94,7 +94,7 @@ TRACE_EVENT(vfio_pci_npu2_mmap,
 #endif /* _TRACE_VFIO_PCI_H */
 
 #undef TRACE_INCLUDE_PATH
-#define TRACE_INCLUDE_PATH .
+#define TRACE_INCLUDE_PATH ../../drivers/vfio/pci
 #undef TRACE_INCLUDE_FILE
 #define TRACE_INCLUDE_FILE trace

Going from the comments in samples/trace_events/trace-events-sample.h, I think both approaches are possible, and I see both used in various places.

Quoted 2 lines
Quoted 2 lines

Personally, I'd prefer Laura's patch, as it doesn't involve hardcoding a path.

Numbering options for clarity:

1)

Quoted 4 lines

ccflags-y += -I$(src)
would add the header search path for all files in drivers/vfio/pci/ whereas only the drivers/vfio/pci/vfio_pci_nvlink2.c needs it.

2)

Quoted 5 lines

CFLAGS_vfio_pci_nvlink2.o += -I$(src)
is a bit better.
However, it is not obvious why this extra header search path is needed until you find vfio_pci_nvlink2.c including trace.h

3)

Quoted 7 lines

#define TRACE_INCLUDE_PATH ../../drivers/vfio/pci
clarifies the intention because the related code is all placed in trace.h

From the comment in include/trace/define_trace.h TRACE_INCLUDE_PATH is relative to include/trace/define_trace.h

In my scan of the tree, the most common solution seems to be 2) as this is essentially recommended in the sample file. 3) is well represented, with much fewer examples of 1), though it might depend how liberally we grep out or examine the use cases. Choice 1) also seems to be the most shotgun approach, adding to the search path for all files.

Quoted 7 lines
Quoted 3 lines

The shotgun approach is always used when compiling out of tree which is what many people do anyway without realizing that there are additional -I. Why do not we make in-tree behave the same way? Thanks,

Are you suggesting bandaging this individual leaf directory, as in choice 1), because it's no worse than what happens for an out-of-tree build anyway,

I suggest making in-tree and out-of-tree behavior equal - both either fail or compile. Just makes sense to me.

Quoted 6 lines

Since out-of-tree adds extra -I, then there should have been a reason for that at the time (before 2005 though). Unfortunately git blame does not say why it was done this way for out-of-tree so imho the safest thing is to add -I for in-tree as well. Or ditch extra -I and do 2) or
3) - this is fine too and can count as an impressive compile optimization, although... look below :)

For the out-of-tree building,
scripts/Makefile.lib adds -I$(srctree)/$(src) and -I$(obj).

In my understanding, -I$(srctree)/$(src) is for including check-in headers from generated C files.

-I$(obj) is for including generated headers from check-in C files.

One example of the correct usage of this is,

crypto/rsa_helper.c (check-in file)

includes

crypto/rsapubkey.asn1.h (generated fle)

Those extra header search paths are unneeded for in-tree building since generated files and check-in files exist in the same tree.

You just happened to find it works for out-of-tree building.

We are talking about how to make include/trace/define_trace.h include drivers/vfio/pci/trace.h

Quoted 8 lines
Quoted 2 lines

The build system cannot (should not)
cater to such a particular case.

Oh well, thanks for the explanation.

Alex, how do we proceed now? I like 2) + comment as in
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/misc/ocxl/Makefile?h=v4.20#n8

It is clear, but
3) TRACE_INCLUDE_PATH solution
is self-documenting.

Quoted 6 lines

better than 3) as relative paths make it dependable on what file actually includes it and it is not clear what is that and what happens if that file moves deeper in the tree (unlikely to happen though).

From the build system PoV, I'd like to discourage people

from adding crappy header search path like this.

Both 2) and 3) have pros and cons.

I just sent 3) patch, but it is the maintainer's call after all.

Quoted 6 lines

--
Alexey

Signature · 2 lines
Best Regards
Masahiro Yamada
#13 lore raw
In reply to Alex Williamson

Alex Williamson <[email protected]> writes:
...

Quoted 9 lines

Numbering options for clarity:

1)

Quoted 4 lines

ccflags-y += -I$(src)
would add the header search path for all files in drivers/vfio/pci/ whereas only the drivers/vfio/pci/vfio_pci_nvlink2.c needs it.

Quoted 10 lines

2)

Quoted 5 lines

CFLAGS_vfio_pci_nvlink2.o += -I$(src)
is a bit better.
However, it is not obvious why this extra header search path is needed until you find vfio_pci_nvlink2.c including trace.h

3)

Quoted 2 lines

#define TRACE_INCLUDE_PATH ../../drivers/vfio/pci
clarifies the intention because the related code is all placed in trace.h

Good summary.

Quoted 7 lines
Quoted 2 lines

From the comment in include/trace/define_trace.h TRACE_INCLUDE_PATH is relative to include/trace/define_trace.h

In my scan of the tree, the most common solution seems to be 2) as this is essentially recommended in the sample file. 3) is well represented, with much fewer examples of 1), though it might depend how liberally we grep out or examine the use cases.

It seems to me that 1 and 2 is overwhelmingly used:

$ git grep -F "#define TRACE_INCLUDE_PATH" | wc -l
133

That counts all definitions of TRACE_INCLUDE_PATH.

$ git grep -F "#define TRACE_INCLUDE_PATH ." | wc -l
122

That's all files using '.', so only 11 locations use the relative path method (3).

Which is unsurprising given that the sample uses '.'.

And people often look at existing code for an example, so they're also going to tend to use '.'.

I agree with Masahiro that adding include paths to the Makefile for this is a bit gross, and method 3 is much more preferable.

Fixing all the existing code to use method 3 would be a good beginner project :)

cheers

#14 lore raw
In reply to Michael Ellerman
Quoted 29 lines · Michael Ellerman
Quoted 10 lines · Laura Abbott

Commit 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver") introduced a trace.h file in the local directory but missed adding the local include path, resulting in compilation failures with tracepoints:

In file included from drivers/vfio/pci/trace.h:102,
from drivers/vfio/pci/vfio_pci_nvlink2.c:29:
./include/trace/define_trace.h:89:42: fatal error: ./trace.h: No such file or directory
#include TRACE_INCLUDE(TRACE_INCLUDE_FILE)

Quoted 8 lines

Fix this by adjusting the include path.

Fixes 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver")

---
I'd still like to echo my sentiment that this should not be a def_bool. We hit this error on our internal testing and we couldn't even turn off the driver until we fixed this.

I assume there's some reason you can't commit a patch to your tree to change it to bool, or turn it off entirely? That would change the SHA which is perhaps reason enough.

In general we have far too many options and most of them never get turned off (or on), so it just creates testing/bug surface for not much benefit. This is one that will probably be turned on in all distro kernels for example.

Quoted 7 lines

But I have no real objection to making it user configurable.

Alex I assume you'll merge this fix via the vfio tree?

cheers

Well now that everything is fixed, we will probably keep it on. My gripe is that when things break like this it's difficult to work around vs. the configuration option at least makes it possible to work around in a fast way. This may also just be a strong preference of mine for working around problems.

Thanks,
Laura

Quoted 11 lines
Quoted 11 lines
drivers/vfio/pci/Makefile +1 -0
index 9662c063a6b1..08d4676a8495 100644
@@ -1,3 +1,4 @@
+ccflags-y                               += -I$(src)
 
  vfio-pci-y := vfio_pci.o vfio_pci_intrs.o vfio_pci_rdwr.o vfio_pci_config.o
  vfio-pci-$(CONFIG_VFIO_PCI_IGD) += vfio_pci_igd.o
Signature · 1 line
2.20.1
Reviewed-by on v1 1/1
#15 lore raw
In reply to Laura Abbott
Quoted 27 lines · On Fri, 4 Jan 2019 11:57:14 -0800 Laura Abbott <[email protected]> wrote:

Commit 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver") introduced a trace.h file in the local directory but missed adding the local include path, resulting in compilation failures with tracepoints:

In file included from drivers/vfio/pci/trace.h:102,
from drivers/vfio/pci/vfio_pci_nvlink2.c:29:
./include/trace/define_trace.h:89:42: fatal error: ./trace.h: No such file or directory
#include TRACE_INCLUDE(TRACE_INCLUDE_FILE)

Fix this by adjusting the include path.

Fixes 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver")

---
I'd still like to echo my sentiment that this should not be a def_bool. We hit this error on our internal testing and we couldn't even turn off the driver until we fixed this.

---
 drivers/vfio/pci/Makefile | 1 +
 1 file changed, 1 insertion(+)
drivers/vfio/pci/Makefile +0 -0
index 9662c063a6b1..08d4676a8495 100644

drivers/vfio/pci/Makefile line 4

Quoted 4 lines
drivers/vfio/pci/Makefile +1 -0
@@ -1,3 +1,4 @@
+ccflags-y                               += -I$(src)
 
 vfio-pci-y := vfio_pci.o vfio_pci_intrs.o vfio_pci_rdwr.o vfio_pci_config.o
 vfio-pci-$(CONFIG_VFIO_PCI_IGD) += vfio_pci_igd.o