* [PATCH kvm-unit-tests] x86: move IDT away from address 0
@ 2020-06-24 16:54 Paolo Bonzini
2020-06-25 6:07 ` Thomas Huth
2020-06-25 18:59 ` Nadav Amit
0 siblings, 2 replies; 8+ messages in thread
From: Paolo Bonzini @ 2020-06-24 16:54 UTC (permalink / raw)
To: kvm; +Cc: mcondotta, Thomas Huth
Address 0 is also used for the SIPI vector (which is probably something worth
changing as well), and now that we call setup_idt very early the SIPI vector
overwrites the first few bytes of the IDT, and in particular the #DE handler.
Fix this for both 32-bit and 64-bit, even though the different form of the
descriptors meant that only 32-bit showed a failure.
Reported-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
x86/cstart.S | 10 +++++++---
x86/cstart64.S | 11 ++++++++++-
2 files changed, 17 insertions(+), 4 deletions(-)
diff --git a/x86/cstart.S b/x86/cstart.S
index 77dc34d..e93dbca 100644
--- a/x86/cstart.S
+++ b/x86/cstart.S
@@ -4,8 +4,6 @@
.globl boot_idt
.global online_cpus
-boot_idt = 0
-
ipi_vector = 0x20
max_cpus = MAX_TEST_CPUS
@@ -30,6 +28,12 @@ i = 0
i = i + 1
.endr
+boot_idt:
+ .rept 256
+ .quad 0
+ .endr
+end_boot_idt:
+
.globl gdt32
gdt32:
.quad 0
@@ -71,7 +75,7 @@ tss:
tss_end:
idt_descr:
- .word 16 * 256 - 1
+ .word end_boot_idt - boot_idt - 1
.long boot_idt
.section .init
diff --git a/x86/cstart64.S b/x86/cstart64.S
index 1ecfbdb..b44d0ae 100644
--- a/x86/cstart64.S
+++ b/x86/cstart64.S
@@ -9,6 +9,8 @@ boot_idt = 0
.globl gdt64_desc
.globl online_cpus
+boot_idt = 0
+
ipi_vector = 0x20
max_cpus = MAX_TEST_CPUS
@@ -51,6 +53,13 @@ ptl5:
.align 4096
+boot_idt:
+ .rept 256
+ .quad 0
+ .quad 0
+ .endr
+end_boot_idt:
+
gdt64_desc:
.word gdt64_end - gdt64 - 1
.quad gdt64
@@ -282,7 +291,7 @@ lvl5:
retq
idt_descr:
- .word 16 * 256 - 1
+ .word end_boot_idt - boot_idt - 1
.quad boot_idt
online_cpus:
--
2.26.2
^ permalink raw reply related [flat|nested] 8+ messages in thread
* Re: [PATCH kvm-unit-tests] x86: move IDT away from address 0
2020-06-24 16:54 [PATCH kvm-unit-tests] x86: move IDT away from address 0 Paolo Bonzini
@ 2020-06-25 6:07 ` Thomas Huth
2020-06-25 18:59 ` Nadav Amit
1 sibling, 0 replies; 8+ messages in thread
From: Thomas Huth @ 2020-06-25 6:07 UTC (permalink / raw)
To: Paolo Bonzini, kvm; +Cc: mcondotta
On 24/06/2020 18.54, Paolo Bonzini wrote:
> Address 0 is also used for the SIPI vector (which is probably something worth
> changing as well), and now that we call setup_idt very early the SIPI vector
> overwrites the first few bytes of the IDT, and in particular the #DE handler.
>
> Fix this for both 32-bit and 64-bit, even though the different form of the
> descriptors meant that only 32-bit showed a failure.
>
> Reported-by: Thomas Huth <thuth@redhat.com>
> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> ---
> x86/cstart.S | 10 +++++++---
> x86/cstart64.S | 11 ++++++++++-
> 2 files changed, 17 insertions(+), 4 deletions(-)
Thanks, this fixes the eventinj test for me!
Tested-by: Thomas Huth <thuth@redhat.com>
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH kvm-unit-tests] x86: move IDT away from address 0
2020-06-24 16:54 [PATCH kvm-unit-tests] x86: move IDT away from address 0 Paolo Bonzini
2020-06-25 6:07 ` Thomas Huth
@ 2020-06-25 18:59 ` Nadav Amit
2020-06-25 19:03 ` Nadav Amit
2020-06-25 19:18 ` Paolo Bonzini
1 sibling, 2 replies; 8+ messages in thread
From: Nadav Amit @ 2020-06-25 18:59 UTC (permalink / raw)
To: Paolo Bonzini; +Cc: kvm, mcondotta, Thomas Huth
> On Jun 24, 2020, at 9:54 AM, Paolo Bonzini <pbonzini@redhat.com> wrote:
>
> Address 0 is also used for the SIPI vector (which is probably something worth
> changing as well), and now that we call setup_idt very early the SIPI vector
> overwrites the first few bytes of the IDT, and in particular the #DE handler.
>
> Fix this for both 32-bit and 64-bit, even though the different form of the
> descriptors meant that only 32-bit showed a failure.
>
> Reported-by: Thomas Huth <thuth@redhat.com>
> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> ---
> x86/cstart.S | 10 +++++++---
> x86/cstart64.S | 11 ++++++++++-
> 2 files changed, 17 insertions(+), 4 deletions(-)
>
> diff --git a/x86/cstart.S b/x86/cstart.S
> index 77dc34d..e93dbca 100644
> --- a/x86/cstart.S
> +++ b/x86/cstart.S
> @@ -4,8 +4,6 @@
> .globl boot_idt
> .global online_cpus
>
> -boot_idt = 0
> -
I think that there is a hidden assumption about the IDT location in
realmode’s test_int(), which this would break:
static void test_int(void)
{
init_inregs(NULL);
boot_idt[11] = 0x1000; /* Store a pointer to address 0x1000 in IDT entry 0x11 */
*(u8 *)(0x1000) = 0xcf; /* 0x1000 contains an IRET instruction */
MK_INSN(int11, "int $0x11\n\t");
exec_in_big_real_mode(&insn_int11);
report("int 1", 0, 1);
}
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH kvm-unit-tests] x86: move IDT away from address 0
2020-06-25 18:59 ` Nadav Amit
@ 2020-06-25 19:03 ` Nadav Amit
2020-06-25 19:18 ` Paolo Bonzini
1 sibling, 0 replies; 8+ messages in thread
From: Nadav Amit @ 2020-06-25 19:03 UTC (permalink / raw)
To: Paolo Bonzini; +Cc: kvm, mcondotta, Thomas Huth
> On Jun 25, 2020, at 11:59 AM, Nadav Amit <nadav.amit@gmail.com> wrote:
>
>> On Jun 24, 2020, at 9:54 AM, Paolo Bonzini <pbonzini@redhat.com> wrote:
>>
>> Address 0 is also used for the SIPI vector (which is probably something worth
>> changing as well), and now that we call setup_idt very early the SIPI vector
>> overwrites the first few bytes of the IDT, and in particular the #DE handler.
>>
>> Fix this for both 32-bit and 64-bit, even though the different form of the
>> descriptors meant that only 32-bit showed a failure.
>>
>> Reported-by: Thomas Huth <thuth@redhat.com>
>> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
>> ---
>> x86/cstart.S | 10 +++++++---
>> x86/cstart64.S | 11 ++++++++++-
>> 2 files changed, 17 insertions(+), 4 deletions(-)
>>
>> diff --git a/x86/cstart.S b/x86/cstart.S
>> index 77dc34d..e93dbca 100644
>> --- a/x86/cstart.S
>> +++ b/x86/cstart.S
>> @@ -4,8 +4,6 @@
>> .globl boot_idt
>> .global online_cpus
>>
>> -boot_idt = 0
>> -
>
> I think that there is a hidden assumption about the IDT location in
> realmode’s test_int(), which this would break.
[ Sorry for the previous wrong quote of my attempt the fix ]
The original offending code:
static void test_int(void)
{
init_inregs(NULL);
*(u32 *)(0x11 * 4) = 0x1000; /* Store a pointer to address 0x1000 in IDT entry 0x11 */
*(u8 *)(0x1000) = 0xcf; /* 0x1000 contains an IRET instruction */
MK_INSN(int11, "int $0x11\n\t");
exec_in_big_real_mode(&insn_int11);
report("int 1", 0, 1);
}
static void test_sti_inhibit(void)
{
init_inregs(NULL);
*(u32 *)(0x73 * 4) = 0x1000; /* Store IRQ 11 handler in the IDT */
*(u8 *)(0x1000) = 0xcf; /* 0x1000 contains an IRET instruction */
MK_INSN(sti_inhibit, "cli\n\t"
"movw $0x200b, %dx\n\t"
"movl $1, %eax\n\t"
"outl %eax, %dx\n\t" /* Set IRQ11 */
"movl $0, %eax\n\t"
"outl %eax, %dx\n\t" /* Clear IRQ11 */
"sti\n\t"
"hlt\n\t");
exec_in_big_real_mode(&insn_sti_inhibit);
report("sti inhibit", ~0, 1);
}
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH kvm-unit-tests] x86: move IDT away from address 0
2020-06-25 18:59 ` Nadav Amit
2020-06-25 19:03 ` Nadav Amit
@ 2020-06-25 19:18 ` Paolo Bonzini
2020-06-26 7:05 ` Paolo Bonzini
1 sibling, 1 reply; 8+ messages in thread
From: Paolo Bonzini @ 2020-06-25 19:18 UTC (permalink / raw)
To: Nadav Amit; +Cc: kvm, mcondotta, Thomas Huth
On 25/06/20 20:59, Nadav Amit wrote:
> I think that there is a hidden assumption about the IDT location in
> realmode’s test_int(), which this would break:
>
> static void test_int(void)
> {
> init_inregs(NULL);
>
> boot_idt[11] = 0x1000; /* Store a pointer to address 0x1000 in IDT entry 0x11 */
> *(u8 *)(0x1000) = 0xcf; /* 0x1000 contains an IRET instruction */
>
> MK_INSN(int11, "int $0x11\n\t");
>
> exec_in_big_real_mode(&insn_int11);
> report("int 1", 0, 1);
> }
Uuuuuuuuuuuuuuuumph... you're right. :( Will send a patch tomorrow.
Paolo
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH kvm-unit-tests] x86: move IDT away from address 0
2020-06-25 19:18 ` Paolo Bonzini
@ 2020-06-26 7:05 ` Paolo Bonzini
2020-06-26 7:06 ` Nadav Amit
0 siblings, 1 reply; 8+ messages in thread
From: Paolo Bonzini @ 2020-06-26 7:05 UTC (permalink / raw)
To: Nadav Amit; +Cc: kvm, mcondotta, Thomas Huth
On 25/06/20 21:18, Paolo Bonzini wrote:
> On 25/06/20 20:59, Nadav Amit wrote:
>> I think that there is a hidden assumption about the IDT location in
>> realmode’s test_int(), which this would break:
>>
>> static void test_int(void)
>> {
>> init_inregs(NULL);
>>
>> boot_idt[11] = 0x1000; /* Store a pointer to address 0x1000 in IDT entry 0x11 */
>> *(u8 *)(0x1000) = 0xcf; /* 0x1000 contains an IRET instruction */
>>
>> MK_INSN(int11, "int $0x11\n\t");
>>
>> exec_in_big_real_mode(&insn_int11);
>> report("int 1", 0, 1);
>> }
>
> Uuuuuuuuuuuuuuuumph... you're right. :( Will send a patch tomorrow.
Actually the IDTR is not reloaded by exec_in_big_real_mode, so this
(while a bit weird) works fine.
Paolo
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH kvm-unit-tests] x86: move IDT away from address 0
2020-06-26 7:05 ` Paolo Bonzini
@ 2020-06-26 7:06 ` Nadav Amit
2020-06-26 7:38 ` Paolo Bonzini
0 siblings, 1 reply; 8+ messages in thread
From: Nadav Amit @ 2020-06-26 7:06 UTC (permalink / raw)
To: Paolo Bonzini; +Cc: kvm, mcondotta, Thomas Huth
> On Jun 26, 2020, at 12:05 AM, Paolo Bonzini <pbonzini@redhat.com> wrote:
>
> On 25/06/20 21:18, Paolo Bonzini wrote:
>> On 25/06/20 20:59, Nadav Amit wrote:
>>> I think that there is a hidden assumption about the IDT location in
>>> realmode’s test_int(), which this would break:
>>>
>>> static void test_int(void)
>>> {
>>> init_inregs(NULL);
>>>
>>> boot_idt[11] = 0x1000; /* Store a pointer to address 0x1000 in IDT entry 0x11 */
>>> *(u8 *)(0x1000) = 0xcf; /* 0x1000 contains an IRET instruction */
>>>
>>> MK_INSN(int11, "int $0x11\n\t");
>>>
>>> exec_in_big_real_mode(&insn_int11);
>>> report("int 1", 0, 1);
>>> }
>>
>> Uuuuuuuuuuuuuuuumph... you're right. :( Will send a patch tomorrow.
>
> Actually the IDTR is not reloaded by exec_in_big_real_mode, so this
> (while a bit weird) works fine.
Err… So it means I need to debug why it does not work for *me*…
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH kvm-unit-tests] x86: move IDT away from address 0
2020-06-26 7:06 ` Nadav Amit
@ 2020-06-26 7:38 ` Paolo Bonzini
0 siblings, 0 replies; 8+ messages in thread
From: Paolo Bonzini @ 2020-06-26 7:38 UTC (permalink / raw)
To: Nadav Amit; +Cc: kvm, mcondotta, Thomas Huth
On 26/06/20 09:06, Nadav Amit wrote:
>> Actually the IDTR is not reloaded by exec_in_big_real_mode, so this
>> (while a bit weird) works fine.
> Err… So it means I need to debug why it does not work for *me*…
Hmm, maybe a dislike for an IDT that is placed above the first MiB of
memory? But I cannot read anything about it in the manuals.
In any case I would accept a patch that switches to the "usual" address
0 IDT in exec_big_real_mode.
Paolo
^ permalink raw reply [flat|nested] 8+ messages in thread
end of thread, other threads:[~2020-06-26 7:38 UTC | newest]
Thread overview: 8+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-06-24 16:54 [PATCH kvm-unit-tests] x86: move IDT away from address 0 Paolo Bonzini
2020-06-25 6:07 ` Thomas Huth
2020-06-25 18:59 ` Nadav Amit
2020-06-25 19:03 ` Nadav Amit
2020-06-25 19:18 ` Paolo Bonzini
2020-06-26 7:05 ` Paolo Bonzini
2020-06-26 7:06 ` Nadav Amit
2020-06-26 7:38 ` Paolo Bonzini
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).