* BUG ? exc_page_fault() was optimized out of fred_hwexc() by gcc with default kernel build option (-O2).
@ 2025-01-18 13:50 Ethan Zhao
2025-01-18 18:34 ` Xin Li
2025-01-19 9:43 ` Ethan Zhao
0 siblings, 2 replies; 4+ messages in thread
From: Ethan Zhao @ 2025-01-18 13:50 UTC (permalink / raw)
To: Xin Li, H. Peter Anvin
Cc: x86, linux-kernel, dave.hansen, tglx, stable, Ethan Zhao
Hi, Xin, Peter
While checking the asm code of arch/x86/entry/entry_fred.o about function fred_hwexc(),
found the code was generated as following :
0000000000000200 <fred_hwexc.constprop.0>:
200: 0f b6 87 a4 00 00 00 movzbl 0xa4(%rdi),%eax
207: 3c 0e cmp $0xe,%al /* match X86_TRAP_PF */
209: 75 05 jne 210 <fred_hwexc.constprop.0+0x10>
20b: e9 00 00 00 00 jmp 210 <fred_hwexc.constprop.0+0x10>
210: 3c 0b cmp $0xb,%al
212: 74 6a je 27e <fred_hwexc.constprop.0+0x7e>
214: 77 17 ja 22d <fred_hwexc.constprop.0+0x2d>
216: 3c 06 cmp $0x6,%al
218: 0f 84 83 00 00 00 je 2a1 <fred_hwexc.constprop.0+0xa1>
21e: 76 29 jbe 249 <fred_hwexc.constprop.0+0x49>
220: 3c 08 cmp $0x8,%al
222: 74 78 je 29c <fred_hwexc.constprop.0+0x9c>
224: 3c 0a cmp $0xa,%al
226: 75 18 jne 240 <fred_hwexc.constprop.0+0x40>
228: e9 00 00 00 00 jmp 22d <fred_hwexc.constprop.0+0x2d>
22d: 3c 11 cmp $0x11,%al
22f: 74 66 je 297 <fred_hwexc.constprop.0+0x97>
231: 76 2c jbe 25f <fred_hwexc.constprop.0+0x5f>
233: 3c 13 cmp $0x13,%al
235: 74 5b je 292 <fred_hwexc.constprop.0+0x92>
237: 3c 15 cmp $0x15,%al
239: 75 1b jne 256 <fred_hwexc.constprop.0+0x56>
23b: e9 00 00 00 00 jmp 240 <fred_hwexc.constprop.0+0x40>
240: 3c 07 cmp $0x7,%al
242: 75 49 jne 28d <fred_hwexc.constprop.0+0x8d>
244: e9 00 00 00 00 jmp 249 <fred_hwexc.constprop.0+0x49>
249: 3c 01 cmp $0x1,%al
24b: 74 3b je 288 <fred_hwexc.constprop.0+0x88>
24d: 3c 05 cmp $0x5,%al
24f: 75 1b jne 26c <fred_hwexc.constprop.0+0x6c>
251: e9 00 00 00 00 jmp 256 <fred_hwexc.constprop.0+0x56>
256: 3c 12 cmp $0x12,%al
258: 75 33 jne 28d <fred_hwexc.constprop.0+0x8d>
25a: e9 00 00 00 00 jmp 25f <fred_hwexc.constprop.0+0x5f>
seems the following calling to exc_page_fault() was optimized out from fred_hwexc() by gcc,
if(likely(regs->fred_ss.vector==X86_TRAP_PF))
returnexc_page_fault(regs,error_code);
gcc version 11.4.0 (Ubuntu 11.4.0-1ubuntu1~22.04)
GNU objdump (GNU Binutils) 2.43
default kernel config.
.config:CONFIG_X86_FRED=y
my understanding, -O2 is the default kernel KBUILD_CFLAGS
So, Are there any workaround needed to make the kernel works with default build ?
or just as Peter said in another loop, manually loading some event bits to make the
over-smart gcc behave normally ？or fall back to -O(ption)0 ?
Any idea, much appreciated !
Thanks,
Ethan
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: BUG ? exc_page_fault() was optimized out of fred_hwexc() by gcc with default kernel build option (-O2).
2025-01-18 13:50 BUG ? exc_page_fault() was optimized out of fred_hwexc() by gcc with default kernel build option (-O2) Ethan Zhao
@ 2025-01-18 18:34 ` Xin Li
2025-01-19 0:42 ` Ethan Zhao
2025-01-19 9:43 ` Ethan Zhao
1 sibling, 1 reply; 4+ messages in thread
From: Xin Li @ 2025-01-18 18:34 UTC (permalink / raw)
To: Ethan Zhao, H. Peter Anvin
Cc: x86, linux-kernel, dave.hansen, tglx, stable, Ethan Zhao
On 1/18/2025 5:50 AM, Ethan Zhao wrote:
> Hi, Xin, Peter
>
> While checking the asm code of arch/x86/entry/entry_fred.o about
> function fred_hwexc(),
> found the code was generated as following :
>
> 0000000000000200 <fred_hwexc.constprop.0>:
> 200: 0f b6 87 a4 00 00 00 movzbl 0xa4(%rdi),%eax
> 207: 3c 0e cmp $0xe,%al /* match X86_TRAP_PF */
> 209: 75 05 jne 210 <fred_hwexc.constprop.0+0x10>
> 20b: e9 00 00 00 00 jmp 210 <fred_hwexc.constprop.0+0x10>
> 210: 3c 0b cmp $0xb,%al
> 212: 74 6a je 27e <fred_hwexc.constprop.0+0x7e>
> 214: 77 17 ja 22d <fred_hwexc.constprop.0+0x2d>
> 216: 3c 06 cmp $0x6,%al
> 218: 0f 84 83 00 00 00 je 2a1 <fred_hwexc.constprop.0+0xa1>
> 21e: 76 29 jbe 249 <fred_hwexc.constprop.0+0x49>
> 220: 3c 08 cmp $0x8,%al
> 222: 74 78 je 29c <fred_hwexc.constprop.0+0x9c>
> 224: 3c 0a cmp $0xa,%al
> 226: 75 18 jne 240 <fred_hwexc.constprop.0+0x40>
> 228: e9 00 00 00 00 jmp 22d <fred_hwexc.constprop.0+0x2d>
> 22d: 3c 11 cmp $0x11,%al
> 22f: 74 66 je 297 <fred_hwexc.constprop.0+0x97>
> 231: 76 2c jbe 25f <fred_hwexc.constprop.0+0x5f>
> 233: 3c 13 cmp $0x13,%al
> 235: 74 5b je 292 <fred_hwexc.constprop.0+0x92>
> 237: 3c 15 cmp $0x15,%al
> 239: 75 1b jne 256 <fred_hwexc.constprop.0+0x56>
> 23b: e9 00 00 00 00 jmp 240 <fred_hwexc.constprop.0+0x40>
> 240: 3c 07 cmp $0x7,%al
> 242: 75 49 jne 28d <fred_hwexc.constprop.0+0x8d>
> 244: e9 00 00 00 00 jmp 249 <fred_hwexc.constprop.0+0x49>
> 249: 3c 01 cmp $0x1,%al
> 24b: 74 3b je 288 <fred_hwexc.constprop.0+0x88>
> 24d: 3c 05 cmp $0x5,%al
> 24f: 75 1b jne 26c <fred_hwexc.constprop.0+0x6c>
> 251: e9 00 00 00 00 jmp 256 <fred_hwexc.constprop.0+0x56>
> 256: 3c 12 cmp $0x12,%al
> 258: 75 33 jne 28d <fred_hwexc.constprop.0+0x8d>
> 25a: e9 00 00 00 00 jmp 25f <fred_hwexc.constprop.0+0x5f>
>
> seems the following calling to exc_page_fault() was optimized out from
> fred_hwexc() by gcc,
>
> if(likely(regs->fred_ss.vector==X86_TRAP_PF))
> returnexc_page_fault(regs,error_code);
>
> gcc version 11.4.0 (Ubuntu 11.4.0-1ubuntu1~22.04)
>
> GNU objdump (GNU Binutils) 2.43
>
>
> default kernel config.
> .config:CONFIG_X86_FRED=y
>
> my understanding, -O2 is the default kernel KBUILD_CFLAGS
> So, Are there any workaround needed to make the kernel works with
> default build ?
> or just as Peter said in another loop, manually loading some event bits
> to make the
> over-smart gcc behave normally ？or fall back to -O(ption)0 ?
>
> Any idea, much appreciated !
This is an optimization done in the original code:
static noinstr void fred_hwexc(struct pt_regs *regs, unsigned long
error_code)
{
/* Optimize for #PF. That's the only exception which matters
performance wise */
if (likely(regs->fred_ss.vector == X86_TRAP_PF))
return exc_page_fault(regs, error_code);
switch (regs->fred_ss.vector) {
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: BUG ? exc_page_fault() was optimized out of fred_hwexc() by gcc with default kernel build option (-O2).
2025-01-18 18:34 ` Xin Li
@ 2025-01-19 0:42 ` Ethan Zhao
0 siblings, 0 replies; 4+ messages in thread
From: Ethan Zhao @ 2025-01-19 0:42 UTC (permalink / raw)
To: Xin Li, H. Peter Anvin
Cc: x86, linux-kernel, dave.hansen, tglx, stable, Ethan Zhao
On 1/19/2025 2:34 AM, Xin Li wrote:
> On 1/18/2025 5:50 AM, Ethan Zhao wrote:
>> Hi, Xin, Peter
>>
>> While checking the asm code of arch/x86/entry/entry_fred.o about
>> function fred_hwexc(),
>> found the code was generated as following :
>>
>> 0000000000000200 <fred_hwexc.constprop.0>:
>> 200: 0f b6 87 a4 00 00 00 movzbl 0xa4(%rdi),%eax
>> 207: 3c 0e cmp $0xe,%al /* match X86_TRAP_PF */
>> 209: 75 05 jne 210
>> <fred_hwexc.constprop.0+0x10>
>> 20b: e9 00 00 00 00 jmp 210
>> <fred_hwexc.constprop.0+0x10>
>> 210: 3c 0b cmp $0xb,%al
>> 212: 74 6a je 27e
>> <fred_hwexc.constprop.0+0x7e>
>> 214: 77 17 ja 22d
>> <fred_hwexc.constprop.0+0x2d>
>> 216: 3c 06 cmp $0x6,%al
>> 218: 0f 84 83 00 00 00 je 2a1
>> <fred_hwexc.constprop.0+0xa1>
>> 21e: 76 29 jbe 249
>> <fred_hwexc.constprop.0+0x49>
>> 220: 3c 08 cmp $0x8,%al
>> 222: 74 78 je 29c
>> <fred_hwexc.constprop.0+0x9c>
>> 224: 3c 0a cmp $0xa,%al
>> 226: 75 18 jne 240
>> <fred_hwexc.constprop.0+0x40>
>> 228: e9 00 00 00 00 jmp 22d
>> <fred_hwexc.constprop.0+0x2d>
>> 22d: 3c 11 cmp $0x11,%al
>> 22f: 74 66 je 297
>> <fred_hwexc.constprop.0+0x97>
>> 231: 76 2c jbe 25f
>> <fred_hwexc.constprop.0+0x5f>
>> 233: 3c 13 cmp $0x13,%al
>> 235: 74 5b je 292
>> <fred_hwexc.constprop.0+0x92>
>> 237: 3c 15 cmp $0x15,%al
>> 239: 75 1b jne 256
>> <fred_hwexc.constprop.0+0x56>
>> 23b: e9 00 00 00 00 jmp 240
>> <fred_hwexc.constprop.0+0x40>
>> 240: 3c 07 cmp $0x7,%al
>> 242: 75 49 jne 28d
>> <fred_hwexc.constprop.0+0x8d>
>> 244: e9 00 00 00 00 jmp 249
>> <fred_hwexc.constprop.0+0x49>
>> 249: 3c 01 cmp $0x1,%al
>> 24b: 74 3b je 288
>> <fred_hwexc.constprop.0+0x88>
>> 24d: 3c 05 cmp $0x5,%al
>> 24f: 75 1b jne 26c
>> <fred_hwexc.constprop.0+0x6c>
>> 251: e9 00 00 00 00 jmp 256
>> <fred_hwexc.constprop.0+0x56>
>> 256: 3c 12 cmp $0x12,%al
>> 258: 75 33 jne 28d
>> <fred_hwexc.constprop.0+0x8d>
>> 25a: e9 00 00 00 00 jmp 25f
>> <fred_hwexc.constprop.0+0x5f>
>>
>> seems the following calling to exc_page_fault() was optimized out
>> from fred_hwexc() by gcc,
>>
>> if(likely(regs->fred_ss.vector==X86_TRAP_PF))
>> returnexc_page_fault(regs,error_code);
>>
>> gcc version 11.4.0 (Ubuntu 11.4.0-1ubuntu1~22.04)
>>
>> GNU objdump (GNU Binutils) 2.43
>>
>>
>> default kernel config.
>> .config:CONFIG_X86_FRED=y
>>
>> my understanding, -O2 is the default kernel KBUILD_CFLAGS
>> So, Are there any workaround needed to make the kernel works with
>> default build ?
>> or just as Peter said in another loop, manually loading some event
>> bits to make the
>> over-smart gcc behave normally ？or fall back to -O(ption)0 ?
>>
>> Any idea, much appreciated !
>
> This is an optimization done in the original code:
>
> static noinstr void fred_hwexc(struct pt_regs *regs, unsigned long
> error_code)
> {
> /* Optimize for #PF. That's the only exception which matters
> performance wise */
> if (likely(regs->fred_ss.vector == X86_TRAP_PF))
The following line code was lost in the asm code after compiling !
Thanks,
Ethan
> return exc_page_fault(regs, error_code);
>
> switch (regs->fred_ss.vector) {
>
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: BUG ? exc_page_fault() was optimized out of fred_hwexc() by gcc with default kernel build option (-O2).
2025-01-18 13:50 BUG ? exc_page_fault() was optimized out of fred_hwexc() by gcc with default kernel build option (-O2) Ethan Zhao
2025-01-18 18:34 ` Xin Li
@ 2025-01-19 9:43 ` Ethan Zhao
1 sibling, 0 replies; 4+ messages in thread
From: Ethan Zhao @ 2025-01-19 9:43 UTC (permalink / raw)
To: Ethan Zhao, Xin Li, H. Peter Anvin
Cc: x86, linux-kernel, dave.hansen, tglx, stable
在 2025/1/18 21:50, Ethan Zhao 写道:
> Hi, Xin, Peter
>
> While checking the asm code of arch/x86/entry/entry_fred.o about
> function fred_hwexc(),
> found the code was generated as following :
>
> 0000000000000200 <fred_hwexc.constprop.0>:
> 200: 0f b6 87 a4 00 00 00 movzbl 0xa4(%rdi),%eax
> 207: 3c 0e cmp $0xe,%al /* match X86_TRAP_PF */
> 209: 75 05 jne 210 <fred_hwexc.constprop.0+0x10>
> 20b: e9 00 00 00 00 jmp 210 <fred_hwexc.constprop.0+0x10>
> 210: 3c 0b cmp $0xb,%al
> 212: 74 6a je 27e <fred_hwexc.constprop.0+0x7e>
> 214: 77 17 ja 22d <fred_hwexc.constprop.0+0x2d>
> 216: 3c 06 cmp $0x6,%al
> 218: 0f 84 83 00 00 00 je 2a1 <fred_hwexc.constprop.0+0xa1>
> 21e: 76 29 jbe 249 <fred_hwexc.constprop.0+0x49>
> 220: 3c 08 cmp $0x8,%al
> 222: 74 78 je 29c <fred_hwexc.constprop.0+0x9c>
> 224: 3c 0a cmp $0xa,%al
> 226: 75 18 jne 240 <fred_hwexc.constprop.0+0x40>
> 228: e9 00 00 00 00 jmp 22d <fred_hwexc.constprop.0+0x2d>
> 22d: 3c 11 cmp $0x11,%al
> 22f: 74 66 je 297 <fred_hwexc.constprop.0+0x97>
> 231: 76 2c jbe 25f <fred_hwexc.constprop.0+0x5f>
> 233: 3c 13 cmp $0x13,%al
> 235: 74 5b je 292 <fred_hwexc.constprop.0+0x92>
> 237: 3c 15 cmp $0x15,%al
> 239: 75 1b jne 256 <fred_hwexc.constprop.0+0x56>
> 23b: e9 00 00 00 00 jmp 240 <fred_hwexc.constprop.0+0x40>
> 240: 3c 07 cmp $0x7,%al
> 242: 75 49 jne 28d <fred_hwexc.constprop.0+0x8d>
> 244: e9 00 00 00 00 jmp 249 <fred_hwexc.constprop.0+0x49>
> 249: 3c 01 cmp $0x1,%al
> 24b: 74 3b je 288 <fred_hwexc.constprop.0+0x88>
> 24d: 3c 05 cmp $0x5,%al
> 24f: 75 1b jne 26c <fred_hwexc.constprop.0+0x6c>
> 251: e9 00 00 00 00 jmp 256 <fred_hwexc.constprop.0+0x56>
> 256: 3c 12 cmp $0x12,%al
> 258: 75 33 jne 28d <fred_hwexc.constprop.0+0x8d>
> 25a: e9 00 00 00 00 jmp 25f <fred_hwexc.constprop.0+0x5f>
>
> seems the following calling to exc_page_fault() was optimized out from
> fred_hwexc() by gcc,
>
> if(likely(regs->fred_ss.vector==X86_TRAP_PF))
> returnexc_page_fault(regs,error_code);
>
> gcc version 11.4.0 (Ubuntu 11.4.0-1ubuntu1~22.04)
>
> GNU objdump (GNU Binutils) 2.43
>
>
> default kernel config.
> .config:CONFIG_X86_FRED=y
>
> my understanding, -O2 is the default kernel KBUILD_CFLAGS
> So, Are there any workaround needed to make the kernel works with
> default build ?
> or just as Peter said in another loop, manually loading some event
> bits to make the
> over-smart gcc behave normally ？or fall back to -O(ption)0 ?
>
> Any idea, much appreciated !
>
>
> Thanks,
> Ethan
>
>
Rebuild the same lastest stable kernel with FRED enabled, default -O2 on RHEL 8.6, got
correct result:
ffffffff820a7430 <fred_hwexc>:
ffffffff820a7430: 55 push %rbp
ffffffff820a7431: 0f b6 87 a4 00 00 00 movzbl 0xa4(%rdi),%eax
ffffffff820a7438: 48 89 e5 mov %rsp,%rbp
ffffffff820a743b: 3c 0e cmp $0xe,%al
ffffffff820a743d: 75 0b jne ffffffff820a744a <fred_hwexc+0x1a>
ffffffff820a743f: e8 4c 50 00 00 callq ffffffff820ac490 <exc_page_fault>
ffffffff820a7444: 5d pop %rbp
gcc version 8.5.0 20210514 (Red Hat 8.5.0-10) (GCC)
whatever CONFIG_JUMP_LABEL=y or CONFIG_JUMP_LABEL is not set.
so the issue seems only about WSL ubuntu toolchain.
Thanks,
Ethan
>
>
--
"firm, enduring, strong, and long-lived"
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2025-01-19 9:43 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2025-01-18 13:50 BUG ? exc_page_fault() was optimized out of fred_hwexc() by gcc with default kernel build option (-O2) Ethan Zhao
2025-01-18 18:34 ` Xin Li
2025-01-19 0:42 ` Ethan Zhao
2025-01-19 9:43 ` Ethan Zhao
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
Powered by JetHome