[v2,0/3] selftests/hid: fix building for older kernels

Message

Justin Stitt Sept. 8, 2023, 10:22 p.m. UTC

Hi, I am sending this series on behalf of myself and Benjamin Tissoires. There
existed an initial n=3 patch series which was later expanded to n=4 and
is now back to n=3 with some fixes added in and rebased against
mainline.

This patch series aims to ensure that the hid/bpf selftests can be built
without errors.

Here's Benjamin's initial cover letter for context:
|  These fixes have been triggered by [0]:
|  basically, if you do not recompile the kernel first, and are
|  running on an old kernel, vmlinux.h doesn't have the required
|  symbols and the compilation fails.
|
|  The tests will fail if you run them on that very same machine,
|  of course, but the binary should compile.
|
|  And while I was sorting out why it was failing, I realized I
|  could do a couple of improvements on the Makefile.
|
|  [0] https://lore.kernel.org/linux-input/[e-mail]/T/#t

Changes from v1 -> v2:
- roll Justin's fix into patch 1/3
- add __attribute__((preserve_access_index)) (thanks Eduard)
- rebased onto mainline (2dde18cd1d8fac735875f2e4987f11817cc0bc2c)
- Link to v1: https://lore.kernel.org/all/[e-mail]/

Link: https://github.com/ClangBuiltLinux/linux/issues/1698
Link: https://github.com/ClangBuiltLinux/continuous-integration2/issues/61
---
Benjamin Tissoires (3):
      selftests/hid: ensure we can compile the tests on kernels pre-6.3
      selftests/hid: do not manually call headers_install
      selftests/hid: force using our compiled libbpf headers

 tools/testing/selftests/hid/Makefile               | 10 ++---
 tools/testing/selftests/hid/progs/hid.c            |  3 --
 .../testing/selftests/hid/progs/hid_bpf_helpers.h  | 49 ++++++++++++++++++++++
 3 files changed, 53 insertions(+), 9 deletions(-)
---
base-commit: 2dde18cd1d8fac735875f2e4987f11817cc0bc2c
change-id: 20230908-kselftest-09-08-56d7f4a8d5c4

Best regards,
--
Justin Stitt <[e-mail]>

Comments

Nick Desaulniers Sept. 8, 2023, 11:02 p.m. UTC | #1

On Fri, Sep 08, 2023 at 10:22:37PM +0000, Justin Stitt wrote:
> Hi, I am sending this series on behalf of myself and Benjamin Tissoires. There
> existed an initial n=3 patch series which was later expanded to n=4 and
> is now back to n=3 with some fixes added in and rebased against
> mainline.
> 
> This patch series aims to ensure that the hid/bpf selftests can be built
> without errors.
> 
> Here's Benjamin's initial cover letter for context:
> |  These fixes have been triggered by [0]:
> |  basically, if you do not recompile the kernel first, and are
> |  running on an old kernel, vmlinux.h doesn't have the required
> |  symbols and the compilation fails.
> |
> |  The tests will fail if you run them on that very same machine,
> |  of course, but the binary should compile.
> |
> |  And while I was sorting out why it was failing, I realized I
> |  could do a couple of improvements on the Makefile.
> |
> |  [0] https://lore.kernel.org/linux-input/[e-mail]/T/#t
> 
> Changes from v1 -> v2:
> - roll Justin's fix into patch 1/3
> - add __attribute__((preserve_access_index)) (thanks Eduard)
> - rebased onto mainline (2dde18cd1d8fac735875f2e4987f11817cc0bc2c)
> - Link to v1: https://lore.kernel.org/all/[e-mail]/
> 
> Link: https://github.com/ClangBuiltLinux/linux/issues/1698
> Link: https://github.com/ClangBuiltLinux/continuous-integration2/issues/61

Thanks to you and Benjamin for sorting all of this out! With this series
applied, I was able to build the hid selftests now without the previous
-Wvisibility diagnostics failing the build.

Tested-by: Nick Desaulniers <[e-mail]> # Build

> ---
> Benjamin Tissoires (3):
>       selftests/hid: ensure we can compile the tests on kernels pre-6.3
>       selftests/hid: do not manually call headers_install
>       selftests/hid: force using our compiled libbpf headers
> 
>  tools/testing/selftests/hid/Makefile               | 10 ++---
>  tools/testing/selftests/hid/progs/hid.c            |  3 --
>  .../testing/selftests/hid/progs/hid_bpf_helpers.h  | 49 ++++++++++++++++++++++
>  3 files changed, 53 insertions(+), 9 deletions(-)
> ---
> base-commit: 2dde18cd1d8fac735875f2e4987f11817cc0bc2c
> change-id: 20230908-kselftest-09-08-56d7f4a8d5c4
> 
> Best regards,
> --
> Justin Stitt <[e-mail]>
>

Justin Stitt Sept. 26, 2023, 7:26 a.m. UTC | #2

Hey all,

Gentle ping on this patch. Looking to get this patch and [1] slated
for 6.7 wherein we can start getting cleaner kselftests builds.

I do not think I am able to successfully run the hid/bpf selftests due
to my kernel version being too low (and an inability to upgrade it as
I'm on a corp rolling release). I'd appreciate some insight on how to
get the tests running or if someone could actually build+run the tests
with this patch applied.

On Sat, Sep 9, 2023 at 7:22 AM Justin Stitt <[e-mail]> wrote:
>
> Hi, I am sending this series on behalf of myself and Benjamin Tissoires. There
> existed an initial n=3 patch series which was later expanded to n=4 and
> is now back to n=3 with some fixes added in and rebased against
> mainline.
>
> This patch series aims to ensure that the hid/bpf selftests can be built
> without errors.
>
> Here's Benjamin's initial cover letter for context:
> |  These fixes have been triggered by [0]:
> |  basically, if you do not recompile the kernel first, and are
> |  running on an old kernel, vmlinux.h doesn't have the required
> |  symbols and the compilation fails.
> |
> |  The tests will fail if you run them on that very same machine,
> |  of course, but the binary should compile.
> |
> |  And while I was sorting out why it was failing, I realized I
> |  could do a couple of improvements on the Makefile.
> |
> |  [0] https://lore.kernel.org/linux-input/[e-mail]/T/#t
>
> Changes from v1 -> v2:
> - roll Justin's fix into patch 1/3
> - add __attribute__((preserve_access_index)) (thanks Eduard)
> - rebased onto mainline (2dde18cd1d8fac735875f2e4987f11817cc0bc2c)
> - Link to v1: https://lore.kernel.org/all/[e-mail]/
>
> Link: https://github.com/ClangBuiltLinux/linux/issues/1698
> Link: https://github.com/ClangBuiltLinux/continuous-integration2/issues/61
> ---
> Benjamin Tissoires (3):
>       selftests/hid: ensure we can compile the tests on kernels pre-6.3
>       selftests/hid: do not manually call headers_install
>       selftests/hid: force using our compiled libbpf headers
>
>  tools/testing/selftests/hid/Makefile               | 10 ++---
>  tools/testing/selftests/hid/progs/hid.c            |  3 --
>  .../testing/selftests/hid/progs/hid_bpf_helpers.h  | 49 ++++++++++++++++++++++
>  3 files changed, 53 insertions(+), 9 deletions(-)
> ---
> base-commit: 2dde18cd1d8fac735875f2e4987f11817cc0bc2c
> change-id: 20230908-kselftest-09-08-56d7f4a8d5c4
>
> Best regards,
> --
> Justin Stitt <[e-mail]>
>

[1]: https://lore.kernel.org/all/[e-mail]/

Thanks
Justin

Benjamin Tissoires Oct. 2, 2023, 2:48 p.m. UTC | #3

On Sep 26 2023, Justin Stitt wrote:
> Hey all,
> 
> Gentle ping on this patch. Looking to get this patch and [1] slated
> for 6.7 wherein we can start getting cleaner kselftests builds.
> 
> I do not think I am able to successfully run the hid/bpf selftests due
> to my kernel version being too low (and an inability to upgrade it as
> I'm on a corp rolling release). I'd appreciate some insight on how to
> get the tests running or if someone could actually build+run the tests
> with this patch applied.

I wanted to apply this series today, but it failed my own CI now with
the enums being already defined:
https://gitlab.freedesktop.org/bentiss/hid/-/jobs/49754306

I'll probably squash the following patch in 1/3, would you mind giving
it a test?

---
From 37feca6c0e84705ad65e621643206c287b63bb0a Mon Sep 17 00:00:00 2001
From: Benjamin Tissoires <[e-mail]>
Date: Mon, 2 Oct 2023 15:37:18 +0200
Subject: [PATCH] fix selftests/hid: ensure we can compile the tests on kernels
 pre-6.3

Signed-off-by: Benjamin Tissoires <[e-mail]>
---
 .../selftests/hid/progs/hid_bpf_helpers.h     | 30 ++++++++++++++++++-
 1 file changed, 29 insertions(+), 1 deletion(-)

diff --git a/tools/testing/selftests/hid/progs/hid_bpf_helpers.h b/tools/testing/selftests/hid/progs/hid_bpf_helpers.h
index ab3b18ba48c4..feed5a991e05 100644
--- a/tools/testing/selftests/hid/progs/hid_bpf_helpers.h
+++ b/tools/testing/selftests/hid/progs/hid_bpf_helpers.h
@@ -5,16 +5,44 @@
 #ifndef __HID_BPF_HELPERS_H
 #define __HID_BPF_HELPERS_H
 
-/* "undefine" structs in vmlinux.h, because we "override" them below */
+/* "undefine" structs and enums in vmlinux.h, because we "override" them below */
 #define hid_bpf_ctx hid_bpf_ctx___not_used
 #define hid_report_type hid_report_type___not_used
 #define hid_class_request hid_class_request___not_used
 #define hid_bpf_attach_flags hid_bpf_attach_flags___not_used
+#define HID_INPUT_REPORT         HID_INPUT_REPORT___not_used
+#define HID_OUTPUT_REPORT        HID_OUTPUT_REPORT___not_used
+#define HID_FEATURE_REPORT       HID_FEATURE_REPORT___not_used
+#define HID_REPORT_TYPES         HID_REPORT_TYPES___not_used
+#define HID_REQ_GET_REPORT       HID_REQ_GET_REPORT___not_used
+#define HID_REQ_GET_IDLE         HID_REQ_GET_IDLE___not_used
+#define HID_REQ_GET_PROTOCOL     HID_REQ_GET_PROTOCOL___not_used
+#define HID_REQ_SET_REPORT       HID_REQ_SET_REPORT___not_used
+#define HID_REQ_SET_IDLE         HID_REQ_SET_IDLE___not_used
+#define HID_REQ_SET_PROTOCOL     HID_REQ_SET_PROTOCOL___not_used
+#define HID_BPF_FLAG_NONE        HID_BPF_FLAG_NONE___not_used
+#define HID_BPF_FLAG_INSERT_HEAD HID_BPF_FLAG_INSERT_HEAD·___not_used
+#define HID_BPF_FLAG_MAX         HID_BPF_FLAG_MAX___not_used
+
 #include "vmlinux.h"
+
 #undef hid_bpf_ctx
 #undef hid_report_type
 #undef hid_class_request
 #undef hid_bpf_attach_flags
+#undef HID_INPUT_REPORT
+#undef HID_OUTPUT_REPORT
+#undef HID_FEATURE_REPORT
+#undef HID_REPORT_TYPES
+#undef HID_REQ_GET_REPORT
+#undef HID_REQ_GET_IDLE
+#undef HID_REQ_GET_PROTOCOL
+#undef HID_REQ_SET_REPORT
+#undef HID_REQ_SET_IDLE
+#undef HID_REQ_SET_PROTOCOL
+#undef HID_BPF_FLAG_NONE
+#undef HID_BPF_FLAG_INSERT_HEAD
+#undef HID_BPF_FLAG_MAX
 
 #include <bpf/bpf_helpers.h>
 #include <bpf/bpf_tracing.h>

Justin Stitt Oct. 3, 2023, 10:34 p.m. UTC | #4

On Mon, Oct 2, 2023 at 7:48 AM Benjamin Tissoires <[e-mail]> wrote:
>
> On Sep 26 2023, Justin Stitt wrote:
> > Hey all,
> >
> > Gentle ping on this patch. Looking to get this patch and [1] slated
> > for 6.7 wherein we can start getting cleaner kselftests builds.
> >
> > I do not think I am able to successfully run the hid/bpf selftests due
> > to my kernel version being too low (and an inability to upgrade it as
> > I'm on a corp rolling release). I'd appreciate some insight on how to
> > get the tests running or if someone could actually build+run the tests
> > with this patch applied.
>
> I wanted to apply this series today, but it failed my own CI now with
> the enums being already defined:
> https://gitlab.freedesktop.org/bentiss/hid/-/jobs/49754306
>
> I'll probably squash the following patch in 1/3, would you mind giving
> it a test?

Works for me with this incantation:
$ make LLVM=1 -j128 ARCH=x86_64 mrproper headers && make LLVM=1 -j128
ARCH=x86_64 -C tools/testing/selftests TARGETS=hid
...
---> BINARY   hid_bpf

Although, the tests expectedly fail.

Looks good to me.

>
> ---
> From 37feca6c0e84705ad65e621643206c287b63bb0a Mon Sep 17 00:00:00 2001
> From: Benjamin Tissoires <[e-mail]>
> Date: Mon, 2 Oct 2023 15:37:18 +0200
> Subject: [PATCH] fix selftests/hid: ensure we can compile the tests on kernels
>  pre-6.3
>
> Signed-off-by: Benjamin Tissoires <[e-mail]>
> ---
>  .../selftests/hid/progs/hid_bpf_helpers.h     | 30 ++++++++++++++++++-
>  1 file changed, 29 insertions(+), 1 deletion(-)
>
> diff --git a/tools/testing/selftests/hid/progs/hid_bpf_helpers.h b/tools/testing/selftests/hid/progs/hid_bpf_helpers.h
> index ab3b18ba48c4..feed5a991e05 100644
> --- a/tools/testing/selftests/hid/progs/hid_bpf_helpers.h
> +++ b/tools/testing/selftests/hid/progs/hid_bpf_helpers.h
> @@ -5,16 +5,44 @@
>  #ifndef __HID_BPF_HELPERS_H
>  #define __HID_BPF_HELPERS_H
>
> -/* "undefine" structs in vmlinux.h, because we "override" them below */
> +/* "undefine" structs and enums in vmlinux.h, because we "override" them below */
>  #define hid_bpf_ctx hid_bpf_ctx___not_used
>  #define hid_report_type hid_report_type___not_used
>  #define hid_class_request hid_class_request___not_used
>  #define hid_bpf_attach_flags hid_bpf_attach_flags___not_used
> +#define HID_INPUT_REPORT         HID_INPUT_REPORT___not_used
> +#define HID_OUTPUT_REPORT        HID_OUTPUT_REPORT___not_used
> +#define HID_FEATURE_REPORT       HID_FEATURE_REPORT___not_used
> +#define HID_REPORT_TYPES         HID_REPORT_TYPES___not_used
> +#define HID_REQ_GET_REPORT       HID_REQ_GET_REPORT___not_used
> +#define HID_REQ_GET_IDLE         HID_REQ_GET_IDLE___not_used
> +#define HID_REQ_GET_PROTOCOL     HID_REQ_GET_PROTOCOL___not_used
> +#define HID_REQ_SET_REPORT       HID_REQ_SET_REPORT___not_used
> +#define HID_REQ_SET_IDLE         HID_REQ_SET_IDLE___not_used
> +#define HID_REQ_SET_PROTOCOL     HID_REQ_SET_PROTOCOL___not_used
> +#define HID_BPF_FLAG_NONE        HID_BPF_FLAG_NONE___not_used
> +#define HID_BPF_FLAG_INSERT_HEAD HID_BPF_FLAG_INSERT_HEAD·___not_used
> +#define HID_BPF_FLAG_MAX         HID_BPF_FLAG_MAX___not_used
> +
>  #include "vmlinux.h"
> +
>  #undef hid_bpf_ctx
>  #undef hid_report_type
>  #undef hid_class_request
>  #undef hid_bpf_attach_flags
> +#undef HID_INPUT_REPORT
> +#undef HID_OUTPUT_REPORT
> +#undef HID_FEATURE_REPORT
> +#undef HID_REPORT_TYPES
> +#undef HID_REQ_GET_REPORT
> +#undef HID_REQ_GET_IDLE
> +#undef HID_REQ_GET_PROTOCOL
> +#undef HID_REQ_SET_REPORT
> +#undef HID_REQ_SET_IDLE
> +#undef HID_REQ_SET_PROTOCOL
> +#undef HID_BPF_FLAG_NONE
> +#undef HID_BPF_FLAG_INSERT_HEAD
> +#undef HID_BPF_FLAG_MAX
>
>  #include <bpf/bpf_helpers.h>
>  #include <bpf/bpf_tracing.h>
> --
> 2.41.0
> ---
>
> Cheers,
> Benjamin
>
> >
> > On Sat, Sep 9, 2023 at 7:22 AM Justin Stitt <[e-mail]> wrote:
> > >
> > > Hi, I am sending this series on behalf of myself and Benjamin Tissoires. There
> > > existed an initial n=3 patch series which was later expanded to n=4 and
> > > is now back to n=3 with some fixes added in and rebased against
> > > mainline.
> > >
> > > This patch series aims to ensure that the hid/bpf selftests can be built
> > > without errors.
> > >
> > > Here's Benjamin's initial cover letter for context:
> > > |  These fixes have been triggered by [0]:
> > > |  basically, if you do not recompile the kernel first, and are
> > > |  running on an old kernel, vmlinux.h doesn't have the required
> > > |  symbols and the compilation fails.
> > > |
> > > |  The tests will fail if you run them on that very same machine,
> > > |  of course, but the binary should compile.
> > > |
> > > |  And while I was sorting out why it was failing, I realized I
> > > |  could do a couple of improvements on the Makefile.
> > > |
> > > |  [0] https://lore.kernel.org/linux-input/[e-mail]/T/#t
> > >
> > > Changes from v1 -> v2:
> > > - roll Justin's fix into patch 1/3
> > > - add __attribute__((preserve_access_index)) (thanks Eduard)
> > > - rebased onto mainline (2dde18cd1d8fac735875f2e4987f11817cc0bc2c)
> > > - Link to v1: https://lore.kernel.org/all/[e-mail]/
> > >
> > > Link: https://github.com/ClangBuiltLinux/linux/issues/1698
> > > Link: https://github.com/ClangBuiltLinux/continuous-integration2/issues/61
> > > ---
> > > Benjamin Tissoires (3):
> > >       selftests/hid: ensure we can compile the tests on kernels pre-6.3
> > >       selftests/hid: do not manually call headers_install
> > >       selftests/hid: force using our compiled libbpf headers
> > >
> > >  tools/testing/selftests/hid/Makefile               | 10 ++---
> > >  tools/testing/selftests/hid/progs/hid.c            |  3 --
> > >  .../testing/selftests/hid/progs/hid_bpf_helpers.h  | 49 ++++++++++++++++++++++
> > >  3 files changed, 53 insertions(+), 9 deletions(-)
> > > ---
> > > base-commit: 2dde18cd1d8fac735875f2e4987f11817cc0bc2c
> > > change-id: 20230908-kselftest-09-08-56d7f4a8d5c4
> > >
> > > Best regards,
> > > --
> > > Justin Stitt <[e-mail]>
> > >
> >
> > [1]: https://lore.kernel.org/all/[e-mail]/
> >
> > Thanks
> > Justin

Thanks
Justin