Skip to contentSkip to conversation

selftests: bpf: modify urandom_read and link it non-statically

PATCH Applied Ivan Vecera 3 participants 7 messages Mar 15, 2019 – Mar 22, 2019

Subsystems (5)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 selftests: bpf: modify urandom_read and link it non-statically

After some experiences I found that urandom_read does not need to be linked statically. When the 'read' syscall call is moved to separate non-inlined function then bpf_get_stackid() is able to find the executable in stack trace and extract its build_id from it.

Signed-off-by Ivan Vecera <ivecera at redhat.com>
---
 tools/testing/selftests/bpf/Makefile       |  2 +-
 tools/testing/selftests/bpf/urandom_read.c | 15 +++++++++++----
 2 files changed, 12 insertions(+), 5 deletions(-)
tools/testing/selftests/bpf/Makefile +1 -1
index 2aed37ea61a4..c33900a8fec0 100644
@@ -69,7 +69,7 @@ TEST_CUSTOM_PROGS = $(OUTPUT)/urandom_read
 all: $(TEST_CUSTOM_PROGS)
 
 $(OUTPUT)/urandom_read: $(OUTPUT)/%: %.c
-	$(CC) -o $@ -static $< -Wl,--build-id
+	$(CC) -o $@ $< -Wl,--build-id
 
 BPFOBJ := $(OUTPUT)/libbpf.a
 
tools/testing/selftests/bpf/urandom_read.c +11 -4
index 9de8b7cb4e6d..db781052758d 100644
@@ -7,11 +7,19 @@
 
 #define BUF_SIZE 256
 
+static __attribute__((noinline))
+void urandom_read(int fd, int count)
+{
+       char buf[BUF_SIZE];
+       int i;
+
+       for (i = 0; i < count; ++i)
+               read(fd, buf, BUF_SIZE);
+}
+
 int main(int argc, char *argv[])
 {
 	int fd = open("/dev/urandom", O_RDONLY);
-	int i;
-	char buf[BUF_SIZE];
 	int count = 4;
 
 	if (fd < 0)
@@ -20,8 +28,7 @@ int main(int argc, char *argv[])
 	if (argc == 2)
 		count = atoi(argv[1]);
 
-	for (i = 0; i < count; ++i)
-		read(fd, buf, BUF_SIZE);
+	urandom_read(fd, count);
 
 	close(fd);
 	return 0;
Signature · 1 line
2.19.2
#1 lore raw
Quoted 4 lines · Ivan Vecera

After some experiences I found that urandom_read does not need to be linked statically. When the 'read' syscall call is moved to separate non-inlined function then bpf_get_stackid() is able to find the executable in stack trace and extract its build_id from it.

But why? Do you have some problems with it being linked statically?

Quoted 59 lines
Signed-off-by Ivan Vecera <ivecera at redhat.com>
---
 tools/testing/selftests/bpf/Makefile       |  2 +-
 tools/testing/selftests/bpf/urandom_read.c | 15 +++++++++++----
 2 files changed, 12 insertions(+), 5 deletions(-)
tools/testing/selftests/bpf/Makefile +1 -1
index 2aed37ea61a4..c33900a8fec0 100644
@@ -69,7 +69,7 @@ TEST_CUSTOM_PROGS = $(OUTPUT)/urandom_read
 all: $(TEST_CUSTOM_PROGS)
 
 $(OUTPUT)/urandom_read: $(OUTPUT)/%: %.c
-	$(CC) -o $@ -static $< -Wl,--build-id
+	$(CC) -o $@ $< -Wl,--build-id
 
 BPFOBJ := $(OUTPUT)/libbpf.a
 
tools/testing/selftests/bpf/urandom_read.c +11 -4
index 9de8b7cb4e6d..db781052758d 100644
@@ -7,11 +7,19 @@
 
 #define BUF_SIZE 256
 
+static __attribute__((noinline))
+void urandom_read(int fd, int count)
+{
+       char buf[BUF_SIZE];
+       int i;
+
+       for (i = 0; i < count; ++i)
+               read(fd, buf, BUF_SIZE);
+}
+
 int main(int argc, char *argv[])
 {
 	int fd = open("/dev/urandom", O_RDONLY);
-	int i;
-	char buf[BUF_SIZE];
 	int count = 4;
 
 	if (fd < 0)
@@ -20,8 +28,7 @@ int main(int argc, char *argv[])
 	if (argc == 2)
 		count = atoi(argv[1]);
 
-	for (i = 0; i < count; ++i)
-		read(fd, buf, BUF_SIZE);
+	urandom_read(fd, count);
 
 	close(fd);
 	return 0;
Signature · 1 line
2.19.2
#2 lore raw
Quoted 7 lines · Stanislav Fomichev
Quoted 4 lines · Ivan Vecera

After some experiences I found that urandom_read does not need to be linked statically. When the 'read' syscall call is moved to separate non-inlined function then bpf_get_stackid() is able to find the executable in stack trace and extract its build_id from it.

But why? Do you have some problems with it being linked statically?

Dependency... you don't need to install static glibc to compile the bpf samples. Shared libc is available everytime.

Ivan

#3 lore raw
Quoted 10 lines · Ivan Vecera
Quoted 7 lines · Stanislav Fomichev
Quoted 4 lines · Ivan Vecera

After some experiences I found that urandom_read does not need to be linked statically. When the 'read' syscall call is moved to separate non-inlined function then bpf_get_stackid() is able to find the executable in stack trace and extract its build_id from it.

But why? Do you have some problems with it being linked statically?

Dependency... you don't need to install static glibc to compile the bpf samples. Shared libc is available everytime.

Oh, the distros that do -devel _and_ -static packages :-)

So your patch essentially adds a call, that leaves a trace on the stack with our build-id. I guess that works as well.

Quoted 2 lines

Ivan

#4 lore raw
Quoted 7 lines · ----- Stanislav Fomichev
Quoted 6 lines · Ivan Vecera
Quoted 5 lines · Stanislav Fomichev
Quoted 4 lines · Ivan Vecera

After some experiences I found that urandom_read does not need to be linked statically. When the 'read' syscall call is moved to separate non-inlined function then bpf_get_stackid() is able to find the executable in stack trace and extract its build_id from it.

Quoted 8 lines
Quoted 4 lines
Quoted 2 lines

But why? Do you have some problems with it being linked statically?

Dependency... you don't need to install static glibc to compile the bpf samples. Shared libc is available everytime.

Oh, the distros that do -devel _and_ -static packages :-)

So your patch essentially adds a call, that leaves a trace on the stack with our build-id. I guess that works as well.

Without that additional call this does not work and build_id selftest fails.

I.

#5 lore raw
Quoted 11 lines · Ivan Vecera
Quoted 9 lines · ----- Stanislav Fomichev
Quoted 8 lines · Ivan Vecera
Quoted 7 lines · Stanislav Fomichev
Quoted 4 lines · Ivan Vecera

After some experiences I found that urandom_read does not need to be linked statically. When the 'read' syscall call is moved to separate non-inlined function then bpf_get_stackid() is able to find the executable in stack trace and extract its build_id from it.

But why? Do you have some problems with it being linked statically?

Quoted 8 lines
Quoted 6 lines
Quoted 2 lines

Dependency... you don't need to install static glibc to compile the bpf samples. Shared libc is available everytime.

Oh, the distros that do -devel _and_ -static packages :-)

So your patch essentially adds a call, that leaves a trace on the stack with our build-id. I guess that works as well.

Without that additional call this does not work and build_id selftest fails.

Oh, yeah, I was just trying to clarify why it fails without -static and why your patch makes it work for non-static :-)
You can put more details in the commit message; you'd have to resubmit whenever net-next/bpf-next opens anyway.

Quoted 2 lines

I.

#6 lore raw
In reply to Ivan Vecera
Quoted 7 lines · Ivan Vecera

After some experiences I found that urandom_read does not need to be linked statically. When the 'read' syscall call is moved to separate non-inlined function then bpf_get_stackid() is able to find the executable in stack trace and extract its build_id from it.

Signed-off-by Ivan Vecera <ivecera at redhat.com>

Applied. Thanks