kvm-unit-test: revert use of RDSEED and RDRAND

PATCH New Matt Weber 3 participants 9 messages Dec 7, 2017 – Jan 9, 2018

Share & tools

Matt Weber

author

2017-12-07 04:36 UTC #0 lore raw

PATCH 1/1 kvm-unit-test: revert use of RDSEED and RDRAND

Removing this commit in buildroot as it causes
build failures when the host binutils isn't
at least 2.23 (2.22.x introduced RDSEED).

The host toolchain is used for x86_64 target
builds where we need to do a 32bit build.

Fixes:
http://autobuild.buildroot.net/results/c39/c3987a3cbd2960b0ff50f872636bdfd8d1a9c820/

Signed-off-by Matthew Weber <[email protected]>

---
 ...002-Revert-x86-vmx-test-RDRAND-and-RDSEED.patch | 66 ++++++++++++++++++++++
 1 file changed, 66 insertions(+)
 create mode 100644 package/kvm-unit-tests/0002-Revert-x86-vmx-test-RDRAND-and-RDSEED.patch

package/kvm-unit-tests/0002-Revert-x86-vmx-test-RDRAND-and-RDSEED.patch new file +66 -0

new file mode 100644
index 0000000000..1d3fb1ca6a
@@ -0,0 +1,66 @@
+From ff6ee5c5345e5a55827d657d8d862237338b6edb Mon Sep 17 00:00:00 2001
+From: Matt Weber <[email protected]>
+Date: Wed, 6 Dec 2017 22:30:04 -0600
+Subject: [PATCH] Revert "x86: vmx: test RDRAND and RDSEED"
+
+Removing this commit in buildroot as it causes
+build failures when the host binutils isn't
+at least 2.23 (2.22.x introduced RDSEED).
+The host toolchain is used for x86_64 target
+builds where we need to do a 32bit build.
+
+Fixes:
+http://autobuild.buildroot.net/results/c39/c3987a3cbd2960b0ff50f872636bdfd8d1a9c820/
+
+This reverts commit a88205d14320d2f681a6c220f1d08711b9ce3885.
+---
+ x86/vmx.h       | 1 -
+ x86/vmx_tests.c | 6 ------
+ 2 files changed, 7 deletions(-)
+
+diff --git a/x86/vmx.h b/x86/vmx.h
+index f0b8776..7a1937e 100644
+--- a/x86/vmx.h
++++ b/x86/vmx.h
+@@ -396,7 +396,6 @@ enum Ctrl1 {
+ 	CPU_URG			= 1ul << 7,
+ 	CPU_VINTD		= 1ul << 9,
+ 	CPU_RDRAND		= 1ul << 11,
+-	CPU_RDSEED		= 1ul << 16,
+ 	CPU_PML                 = 1ul << 17,
+ };
+ 
+diff --git a/x86/vmx_tests.c b/x86/vmx_tests.c
+index e8c97f2..3619e0c 100644
+--- a/x86/vmx_tests.c
++++ b/x86/vmx_tests.c
+@@ -772,8 +772,6 @@ asm(
+ 	"insn_sldt: sldt %ax;ret\n\t"
+ 	"insn_lldt: xor %eax, %eax; lldt %ax;ret\n\t"
+ 	"insn_str: str %ax;ret\n\t"
+-	"insn_rdrand: rdrand %rax;ret\n\t"
+-	"insn_rdseed: rdseed %rax;ret\n\t"
+ );
+ extern void insn_hlt();
+ extern void insn_invlpg();
+@@ -798,8 +796,6 @@ extern void insn_lldt();
+ extern void insn_str();
+ extern void insn_cpuid();
+ extern void insn_invd();
+-extern void insn_rdrand();
+-extern void insn_rdseed();
+ 
+ u32 cur_insn;
+ u64 cr3;
+@@ -855,8 +851,6 @@ static struct insn_table insn_table[] = {
+ 	{"DESC_TABLE (LLDT)", CPU_DESC_TABLE, insn_lldt, INSN_CPU1, 47, 0, 0, 0},
+ 	{"DESC_TABLE (STR)", CPU_DESC_TABLE, insn_str, INSN_CPU1, 47, 0, 0, 0},
+ 	/* LTR causes a #GP if done with a busy selector, so it is not tested.  */
+-	{"RDRAND", CPU_RDRAND, insn_rdrand, INSN_CPU1, VMX_RDRAND, 0, 0, 0},
+-	{"RDSEED", CPU_RDSEED, insn_rdseed, INSN_CPU1, VMX_RDSEED, 0, 0, 0},
+ 	// Instructions always trap
+ 	{"CPUID", 0, insn_cpuid, INSN_ALWAYS_TRAP, 10, 0, 0, 0},
+ 	{"INVD", 0, insn_invd, INSN_ALWAYS_TRAP, 13, 0, 0, 0},
+-- 
+2.14.2
+

Signature · 1 line

2.14.2

Thomas Petazzoni

2017-12-14 08:44 UTC #1 lore raw

PATCH 1/1 kvm-unit-test: revert use of RDSEED and RDRAND

Hello,

Quoted 11 lines · Matt Weber

Removing this commit in buildroot as it causes
build failures when the host binutils isn't
at least 2.23 (2.22.x introduced RDSEED).
The host toolchain is used for x86_64 target
builds where we need to do a 32bit build.
Fixes:
http://autobuild.buildroot.net/results/c39/c3987a3cbd2960b0ff50f872636bdfd8d1a9c820/
Signed-off-by Matthew Weber <[email protected]>

What is the long term solution for this? Has the problem been reported upstream?

Quoted 11 lines

package/kvm-unit-tests/0002-Revert-x86-vmx-test-RDRAND-and-RDSEED.patch new file +5 -0

new file mode 100644
index 0000000000..1d3fb1ca6a
@@ -0,0 +1,66 @@
+From ff6ee5c5345e5a55827d657d8d862237338b6edb Mon Sep 17 00:00:00 2001
+From: Matt Weber <[email protected]>
+Date: Wed, 6 Dec 2017 22:30:04 -0600
+Subject: [PATCH] Revert "x86: vmx: test RDRAND and RDSEED"
+

package/kvm-unit-tests/0002-Revert-x86-vmx-test-RDRAND-and-RDSEED.patch line 15

Quoted 10 lines

package/kvm-unit-tests/0002-Revert-x86-vmx-test-RDRAND-and-RDSEED.patch new file +10 -0

+Removing this commit in buildroot as it causes
+build failures when the host binutils isn't
+at least 2.23 (2.22.x introduced RDSEED).
+The host toolchain is used for x86_64 target
+builds where we need to do a 32bit build.
+
+Fixes:
+http://autobuild.buildroot.net/results/c39/c3987a3cbd2960b0ff50f872636bdfd8d1a9c820/
+
+This reverts commit a88205d14320d2f681a6c220f1d08711b9ce3885.

Missing SoB.

Thanks,

Thomas

Signature · 3 lines

Thomas Petazzoni, CTO, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com

Matthew Weber

author

2017-12-15 20:41 UTC #2 lore raw

PATCH 1/1 kvm-unit-test: revert use of RDSEED and RDRAND

Thomas,

Quoted 11 lines · Thomas Petazzoni

Hello,
Quoted 7 lines · Matt Weber
Removing this commit in buildroot as it causes
build failures when the host binutils isn't
at least 2.23 (2.22.x introduced RDSEED).
The host toolchain is used for x86_64 target
builds where we need to do a 32bit build.

Quoted 7 lines

Quoted 4 lines
Fixes:
http://autobuild.buildroot.net/results/c39/c3987a3cbd2960b0ff50f872636bdfd8d1a9c820/
Signed-off-by Matthew Weber <[email protected]>
What is the long term solution for this? Has the problem been reported upstream?

It isn't really a bug upstream. I think the long term fix would be a check for a minimum binutils version. Can you think of another case I could use as an example where a evolution in the binutils has required a minimum version?

Quoted 16 lines

package/kvm-unit-tests/0002-Revert-x86-vmx-test-RDRAND-and-RDSEED.patch new file +9 -0

new file mode 100644
index 0000000000..1d3fb1ca6a
@@ -0,0 +1,66 @@
+From ff6ee5c5345e5a55827d657d8d862237338b6edb Mon Sep 17 00:00:00 2001
+From: Matt Weber <[email protected]>
+Date: Wed, 6 Dec 2017 22:30:04 -0600
+Subject: [PATCH] Revert "x86: vmx: test RDRAND and RDSEED"
+
+Removing this commit in buildroot as it causes
+build failures when the host binutils isn't
+at least 2.23 (2.22.x introduced RDSEED).
+The host toolchain is used for x86_64 target

Quoted 8 lines

package/kvm-unit-tests/0002-Revert-x86-vmx-test-RDRAND-and-RDSEED.patch line 15
Quoted 6 lines
package/kvm-unit-tests/0002-Revert-x86-vmx-test-RDRAND-and-RDSEED.patch new file +6 -0
+builds where we need to do a 32bit build.
+
+Fixes:
+http://autobuild.buildroot.net/results/c39/c3987a3cbd2960b0ff50f872636bdfd8d1a9c820/
+
+This reverts commit a88205d14320d2f681a6c220f1d08711b9ce3885.
Missing SoB.

Opps, noted.

Matt

Kees van Unen

2017-12-15 20:41 UTC #3 lore raw

PATCH 1/1 kvm-unit-test: revert use of RDSEED and RDRAND

________________________________________
Van: buildrootNamensMatthew Weber
Verzonden: vrijdag 15 december 2017 21:41:05 (UTC+01:00) Amsterdam, Berlijn, Bern, Rome, Stockholm, Wenen
Aan: Thomas Petazzoni
CC: buildroot
Onderwerp: Re: [Buildroot] [PATCH] kvm-unit-test: revert use of RDSEED and RDRAND

Thomas,

Quoted 11 lines · Thomas Petazzoni

Hello,
Quoted 7 lines · Matt Weber
Removing this commit in buildroot as it causes
build failures when the host binutils isn't
at least 2.23 (2.22.x introduced RDSEED).
The host toolchain is used for x86_64 target
builds where we need to do a 32bit build.

Quoted 7 lines

Quoted 4 lines
Fixes:
http://autobuild.buildroot.net/results/c39/c3987a3cbd2960b0ff50f872636bdfd8d1a9c820/
Signed-off-by Matthew Weber <[email protected]>
What is the long term solution for this? Has the problem been reported upstream?

Quoted 16 lines

package/kvm-unit-tests/0002-Revert-x86-vmx-test-RDRAND-and-RDSEED.patch new file +9 -0

new file mode 100644
index 0000000000..1d3fb1ca6a
@@ -0,0 +1,66 @@
+From ff6ee5c5345e5a55827d657d8d862237338b6edb Mon Sep 17 00:00:00 2001
+From: Matt Weber <[email protected]>
+Date: Wed, 6 Dec 2017 22:30:04 -0600
+Subject: [PATCH] Revert "x86: vmx: test RDRAND and RDSEED"
+
+Removing this commit in buildroot as it causes
+build failures when the host binutils isn't
+at least 2.23 (2.22.x introduced RDSEED).
+The host toolchain is used for x86_64 target

Quoted 8 lines

package/kvm-unit-tests/0002-Revert-x86-vmx-test-RDRAND-and-RDSEED.patch line 15
Quoted 6 lines
package/kvm-unit-tests/0002-Revert-x86-vmx-test-RDRAND-and-RDSEED.patch new file +6 -0
+builds where we need to do a 32bit build.
+
+Fixes:
+http://autobuild.buildroot.net/results/c39/c3987a3cbd2960b0ff50f872636bdfd8d1a9c820/
+
+This reverts commit a88205d14320d2f681a6c220f1d08711b9ce3885.
Missing SoB.

Opps, noted.

Matt
_______________________________________________
buildroot mailing list
buildroot at busybox.net
http://lists.busybox.net/mailman/listinfo/buildroot

Thomas Petazzoni

2017-12-18 07:49 UTC #4 lore raw

PATCH 1/1 kvm-unit-test: revert use of RDSEED and RDRAND

In reply to Matthew Weber

Hello,

Quoted 2 lines · Matthew Weber

It isn't really a bug upstream. I think the long term fix would be a check for a minimum binutils version.

Or kvm-unit-tests could do a configure test to verify if rdseed/rdrand are supported, and only use them if so.

Quoted 2 lines

Can you think of another case I could use as an example where a evolution in the binutils has required a minimum version?

I don't think we had such a case so far. We do have BR2_TOOLCHAIN_BINUTILS_HAS_BUG_xyz, but I don't think we have had version dependencies for binutils.

Thomas

Signature · 3 lines

Thomas Petazzoni, CTO, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com

Thomas Petazzoni

2018-01-03 22:27 UTC #5 lore raw

PATCH 1/1 kvm-unit-test: revert use of RDSEED and RDRAND

Hello,

Quoted 7 lines · Thomas Petazzoni

Quoted 2 lines · Matthew Weber
It isn't really a bug upstream. I think the long term fix would be a check for a minimum binutils version.
Or kvm-unit-tests could do a configure test to verify if rdseed/rdrand are supported, and only use them if so.

Any chance you could submit a patch implementing this?

Thanks!

Thomas

Signature · 3 lines

Thomas Petazzoni, CTO, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com

Matthew Weber

author

2018-01-03 22:42 UTC #6 lore raw

PATCH 1/1 kvm-unit-test: revert use of RDSEED and RDRAND

Thomas,

Quoted 6 lines · Thomas Petazzoni

Hello,
Quoted 2 lines · Thomas Petazzoni
On Fri, 15 Dec 2017 14:41:05 -0600, Matthew Weber wrote:

Quoted 8 lines

Quoted 5 lines
Quoted 2 lines
It isn't really a bug upstream. I think the long term fix would be a check for a minimum binutils version.
Or kvm-unit-tests could do a configure test to verify if rdseed/rdrand are supported, and only use them if so.
Any chance you could submit a patch implementing this?

Yeah, I've done the research but was still looking at how. Here's where I was at below.......

I'd default to disable because we don't have any way to tell what the end processor is for execution of what we build. Worst case we could use the cpu type and enable a new configure option to turn this feature on/off. I can suggest that to the mailing list and maybe they'd be ok with it defaulting to on.

Matt

Thomas Petazzoni

2018-01-04 08:36 UTC #7 lore raw

PATCH 1/1 kvm-unit-test: revert use of RDSEED and RDRAND

Hello,

Quoted 8 lines · Matthew Weber

Yeah, I've done the research but was still looking at how. Here's where I was at below.......
I'd default to disable because we don't have any way to tell what the end processor is for execution of what we build. Worst case we could use the cpu type and enable a new configure option to turn this feature on/off. I can suggest that to the mailing list and maybe they'd be ok with it defaulting to on.

Well, gcc/binutils will barf out with an error if the selected processor variant (via mcpu) doesn't support the instruction.

Of course, if you specify a bogus mcpu flag that doesn't match your target CPU, then it won't run. But well, passing a bogus mcpu is already going to generate code that won't run: if you build for core-i7 and run on your old i486, the code won't run.

Best regards,

Thomas

Signature · 3 lines

Thomas Petazzoni, CTO, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com

Matthew Weber

author

2018-01-09 16:09 UTC #8 lore raw

PATCH 1/1 kvm-unit-test: revert use of RDSEED and RDRAND

Thomas,

Quoted 8 lines · Thomas Petazzoni

Hello,
Quoted 3 lines · Matthew Weber
Yeah, I've done the research but was still looking at how. Here's where I was at below.......

Quoted 9 lines

Quoted 5 lines
I'd default to disable because we don't have any way to tell what the end processor is for execution of what we build. Worst case we could use the cpu type and enable a new configure option to turn this feature on/off. I can suggest that to the mailing list and maybe they'd be ok with it defaulting to on.
Well, gcc/binutils will barf out with an error if the selected processor variant (via mcpu) doesn't support the instruction.

Superseded by:
https://patchwork.ozlabs.org/patch/857418/