public inbox for kvm@vger.kernel.org
 help / color / mirror / Atom feed
From: "Shukla, Manali" <mashukla@amd.com>
To: Aaron Lewis <aaronlewis@google.com>,
	Manali Shukla <manali.shukla@amd.com>
Cc: pbonzini@redhat.com, kvm@vger.kernel.org, seanjc@google.com
Subject: Re: [kvm-unit-tests PATCH 3/3] x86: nSVM: Add an exception test framework and tests
Date: Thu, 17 Feb 2022 08:56:25 +0530	[thread overview]
Message-ID: <18489ffd-c3bc-2f0c-38cf-9faa74cf3363@amd.com> (raw)
In-Reply-To: <CAAAPnDH6y6pFG+Mw_JCYYi9rome0d0+Q4UTLK3KoBzREvkJwqw@mail.gmail.com>



On 2/15/2022 1:50 AM, Aaron Lewis wrote:
>> +static void svm_l2_nm_test(struct svm_test *svm)
>> +{
>> +    write_cr0(read_cr0() | X86_CR0_TS);
>> +    asm volatile("fnop");
>> +}
>> +
>> +static void svm_l2_of_test(struct svm_test *svm)
>> +{
>> +    struct far_pointer32 fp = {
>> +        .offset = (uintptr_t)&&into,
>> +        .selector = KERNEL_CS32,
>> +    };
>> +    uintptr_t rsp;
>> +
>> +    asm volatile ("mov %%rsp, %0" : "=r"(rsp));
>> +
>> +    if (fp.offset != (uintptr_t)&&into) {
>> +        printf("Codee address too high.\n");
> 
> Nit: Code
> 
>> +        return;
>> +    }
>> +
>> +    if ((u32)rsp != rsp) {
>> +        printf("Stack address too high.\n");
>> +    }
>> +
>> +    asm goto("lcall *%0" : : "m" (fp) : "rax" : into);
>> +    return;
>> +into:
>> +    asm volatile (".code32;"
>> +            "movl $0x7fffffff, %eax;"
>> +            "addl %eax, %eax;"
>> +            "into;"
>> +            "lret;"
>> +            ".code64");
>> +    __builtin_unreachable();
>> +}
>> +
> 
>> +static void svm_l2_ac_test(struct svm_test *test)
>> +{
>> +    bool hit_ac = false;
>> +
>> +    write_cr0(read_cr0() | X86_CR0_AM);
>> +    write_rflags(read_rflags() | X86_EFLAGS_AC);
>> +
>> +    run_in_user(usermode_callback, AC_VECTOR, 0, 0, 0, 0, &hit_ac);
>> +
>> +    report(hit_ac, "Usermode #AC handled in L2");
>> +    vmmcall();
>> +}
>> +
>> +static void svm_ac_init(void)
>> +{
>> +    set_user_mask_all(phys_to_virt(read_cr3()), PAGE_LEVEL);
>> +}
>> +
>> +static void svm_ac_uninit(void)
>> +{
>> +    clear_user_mask_all(phys_to_virt(read_cr3()), PAGE_LEVEL);
>> +}
>> +
>> +struct svm_exception_test {
>> +    u8 vector;
>> +    void (*guest_code)(struct svm_test*);
>> +    void (*init_test)(void);
>> +    void (*uninit_test)(void);
>> +};
>> +
>> +struct svm_exception_test svm_exception_tests[] = {
>> +    { GP_VECTOR, svm_l2_gp_test },
>> +    { UD_VECTOR, svm_l2_ud_test },
>> +    { DE_VECTOR, svm_l2_de_test },
>> +    { BP_VECTOR, svm_l2_bp_test },
>> +    { NM_VECTOR, svm_l2_nm_test },
>> +    { OF_VECTOR, svm_l2_of_test },
>> +    { DB_VECTOR, svm_l2_db_test },
>> +    { AC_VECTOR, svm_l2_ac_test, svm_ac_init, svm_ac_uninit },
>> +};
> 
> If you set and clear PT_USER_MASK in svm_l2_ac_test() before calling
> into userspace you can remove init_test and uninit_test from the
> framework all together.  That will simplify the code.
> 
If clear user mask is called after userspace code, when #AC exception is 
intercepted by L1, the control directly goes to L1 and it does not reach 
clear_user_mask_all() function (called after user space code function run_in_user()).

That is why I have added init_test and uninit_test function

> Further, it would be nice to then hoist this framework and the one in
> vmx into a common x86 file, but looking at this that may be something
> to think about in the future.  There would have to be wrappers when
> interacting with the vmc{s,b} and macros at the very least.
> 

Yeah we can think of this in future.

>> +
>> +static u8 svm_exception_test_vector;
>> +
>> +static void svm_exception_handler(struct ex_regs *regs)
>> +{
>> +    report(regs->vector == svm_exception_test_vector,
>> +            "Handling %s in L2's exception handler",
>> +            exception_mnemonic(svm_exception_test_vector));
>> +    vmmcall();
>> +}
>> +

-Manali

  reply	other threads:[~2022-02-17  3:26 UTC|newest]

Thread overview: 13+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-02-07  5:11 [kvm-unit-tests PATCH 0/3] nSVM: Add testing for routing L2 exceptions Manali Shukla
2022-02-07  5:12 ` [kvm-unit-tests PATCH 1/3] x86: Add routines to set/clear PT_USER_MASK for all pages Manali Shukla
2022-02-14 19:30   ` Sean Christopherson
2022-02-17  3:55     ` Shukla, Manali
2022-02-17 14:34       ` Aaron Lewis
2022-02-20  4:42         ` Shukla, Manali
2022-02-17 15:20       ` Sean Christopherson
2022-02-20  5:35         ` Shukla, Manali
2022-02-07  5:12 ` [kvm-unit-tests PATCH 2/3] x86: Make exception_mnemonic() visible to the tests Manali Shukla
2022-02-07  5:12 ` [kvm-unit-tests PATCH 3/3] x86: nSVM: Add an exception test framework and tests Manali Shukla
2022-02-14 20:20   ` Aaron Lewis
2022-02-17  3:26     ` Shukla, Manali [this message]
2022-02-17 14:46       ` Aaron Lewis

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=18489ffd-c3bc-2f0c-38cf-9faa74cf3363@amd.com \
    --to=mashukla@amd.com \
    --cc=aaronlewis@google.com \
    --cc=kvm@vger.kernel.org \
    --cc=manali.shukla@amd.com \
    --cc=pbonzini@redhat.com \
    --cc=seanjc@google.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox