[PATCH] x86/hweight: Don't clobber %rdi

[ville . syrjala]

From: Ville SyrjÃlà <ville.syrjala@xxxxxxxxxxxxxxx>

The caller expects %rdi to remain intact, push+pop it make that happen.

Fixes the following kind of explosions on my core2duo machine when
trying to reboot or shut down:

general protection fault: 0000 [#1] PREEMPT SMP
Modules linked in: i915 i2c_algo_bit drm_kms_helper cfbfillrect syscopyarea cfbimgblt sysfillrect sysimgblt fb_sys_fops cfbcopyarea drm netconsole configfs binfmt_misc iTCO_wdt psmouse pcspkr snd_hda_codec_idt e100 coretemp hwmon snd_hda_codec_generic i2c_i801 mii i2c_smbus lpc_ich mfd_core snd_hda_intel uhci_hcd snd_hda_codec snd_hwdep snd_hda_core ehci_pci 8250 ehci_hcd snd_pcm 8250_base usbcore evdev serial_core usb_common parport_pc parport snd_timer snd soundcore
CPU: 0 PID: 3070 Comm: reboot Not tainted 4.8.0-rc1-perf-dirty #69
Hardware name:                  /D946GZIS, BIOS TS94610J.86A.0087.2007.1107.1049 11/07/2007
task: ffff88012a0b4080 task.stack: ffff880123850000
RIP: 0010:[<ffffffff81003c92>]  [<ffffffff81003c92>] x86_perf_event_update+0x52/0xc0
RSP: 0018:ffff880123853b60  EFLAGS: 00010087
RAX: 0000000000000001 RBX: ffff88012fc0a3c0 RCX: 000000000000001e
RDX: 0000000000000000 RSI: 0000000040000000 RDI: ffff88012b014800
RBP: ffff880123853b88 R08: ffffffffffffffff R09: 0000000000000000
R10: ffffea0004a012c0 R11: ffffea0004acedc0 R12: ffffffff80000001
R13: ffff88012b0149c0 R14: ffff88012b014800 R15: 0000000000000018
FS:  00007f8b155cd700(0000) GS:ffff88012fc00000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007f8b155f5000 CR3: 000000012a2d7000 CR4: 00000000000006f0
Stack:
 ffff88012fc0a3c0 ffff88012b014800 0000000000000004 0000000000000001
 ffff88012fc1b750 ffff880123853bb0 ffffffff81003d59 ffff88012b014800
 ffff88012fc0a3c0 ffff88012b014800 ffff880123853bd8 ffffffff81003e13
Call Trace:
 [<ffffffff81003d59>] x86_pmu_stop+0x59/0xd0
 [<ffffffff81003e13>] x86_pmu_del+0x43/0x140
 [<ffffffff8111705d>] event_sched_out.isra.105+0xbd/0x260
 [<ffffffff8111738d>] __perf_remove_from_context+0x2d/0xb0
 [<ffffffff8111745d>] __perf_event_exit_context+0x4d/0x70
 [<ffffffff810c8826>] generic_exec_single+0xb6/0x140
 [<ffffffff81117410>] ? __perf_remove_from_context+0xb0/0xb0
 [<ffffffff81117410>] ? __perf_remove_from_context+0xb0/0xb0
 [<ffffffff810c898f>] smp_call_function_single+0xdf/0x140
 [<ffffffff81113d27>] perf_event_exit_cpu_context+0x87/0xc0
 [<ffffffff81113d73>] perf_reboot+0x13/0x40
 [<ffffffff8107578a>] notifier_call_chain+0x4a/0x70
 [<ffffffff81075ad7>] __blocking_notifier_call_chain+0x47/0x60
 [<ffffffff81075b06>] blocking_notifier_call_chain+0x16/0x20
 [<ffffffff81076a1d>] kernel_restart_prepare+0x1d/0x40
 [<ffffffff81076ae2>] kernel_restart+0x12/0x60
 [<ffffffff81076d56>] SYSC_reboot+0xf6/0x1b0
 [<ffffffff811a823c>] ? mntput_no_expire+0x2c/0x1b0
 [<ffffffff811a83e4>] ? mntput+0x24/0x40
 [<ffffffff811894fc>] ? __fput+0x16c/0x1e0
 [<ffffffff811895ae>] ? ____fput+0xe/0x10
 [<ffffffff81072fc3>] ? task_work_run+0x83/0xa0
 [<ffffffff81001623>] ? exit_to_usermode_loop+0x53/0xc0
 [<ffffffff8100105a>] ? trace_hardirqs_on_thunk+0x1a/0x1c
 [<ffffffff81076e6e>] SyS_reboot+0xe/0x10
 [<ffffffff814c4ba5>] entry_SYSCALL_64_fastpath+0x18/0xa3
Code: 7c 4c 8d af c0 01 00 00 49 89 fe eb 10 48 09 c2 4c 89 e0 49 0f b1 55 00 4c 39 e0 74 35 4d 8b a6 c0 01 00 00 41 8b 8e 60 01 00 00 <0f> 33 8b 35 6e 02 8c 00 48 c1 e2 20 85 f6 7e d2 48 89 d3 89 cf
RIP  [<ffffffff81003c92>] x86_perf_event_update+0x52/0xc0
 RSP <ffff880123853b60>
---[ end trace 7ec95181faf211be ]---
note: reboot[3070] exited with preempt_count 2

Cc: Borislav Petkov <bp@xxxxxxx>
Cc: H. Peter Anvin <hpa@xxxxxxxxx>
Cc: Andy Lutomirski <luto@xxxxxxxxxxxxxx>
Cc: Brian Gerst <brgerst@xxxxxxxxx>
Cc: Denys Vlasenko <dvlasenk@xxxxxxxxxx>
Cc: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>
Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
Cc: Ingo Molnar <mingo@xxxxxxxxxx>
Fixes: f5967101e9de ("x86/hweight: Get rid of the special calling convention")
Signed-off-by: Ville SyrjÃlà <ville.syrjala@xxxxxxxxxxxxxxx>
---
 arch/x86/lib/hweight.S | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/arch/x86/lib/hweight.S b/arch/x86/lib/hweight.S
index 02de3d74d2c5..8a602a1e404a 100644
--- a/arch/x86/lib/hweight.S
+++ b/arch/x86/lib/hweight.S
@@ -35,6 +35,7 @@ ENDPROC(__sw_hweight32)
 
 ENTRY(__sw_hweight64)
 #ifdef CONFIG_X86_64
+	pushq   %rdi
 	pushq   %rdx
 
 	movq    %rdi, %rdx                      # w -> t
@@ -60,6 +61,7 @@ ENTRY(__sw_hweight64)
 	shrq    $56, %rax                       # w = w_tmp >> 56
 
 	popq    %rdx
+	popq    %rdi
 	ret
 #else /* CONFIG_X86_32 */
 	/* We're getting an u64 arg in (%eax,%edx): unsigned long hweight64(__u64 w) */
-- 
2.7.4