Double Faults (zh-CN)

[giscus[bot]]

Double Faults (zh-CN)

在这篇文章中，我们会探索 double fault 异常的细节，它的触发条件是调用错误处理函数失败。通过捕获该异常，我们可以阻止致命的 triple faults 异常导致系统重启。为了尽可能避免 triple faults ，我们会在一个独立的内核栈配置 中断栈表 来捕捉 double fault…

https://os.phil-opp.com/zh-CN/double-fault-exceptions/

[nliao6622]

Well, first, this is a very good tutorial.

A quick question: when I try to repeat the page fault at 0xdeadbeef in very first example on macbook, I got this following message:

"panic at 'misaligned pointer dereference: address must be a multiple of 0x8 but is 0xdeadbeef' src/main.rs:21:9"

More interestingly, the fault is the same even after I defined the double_fault handler. Looks the double_fault_handler is never triggered.

My platform is Macbook Big Sur Intel based. What could be the like cause of the problem?

Thanks!

[ghost]

I had the same issue. The address has to be a multiple of 8, which 0xdeadbeef (3735928559 in decimal)
is not. Try using 0xdeadbef0 (3735928560). This might be helpful http://junyelee.blogspot.com/2019/01/memory-alignment-issue.html

[SPuntte]

I stumbled upon this as well. Indeed, the address 0xdeadbeef is not properly aligned for a *mut u64. Rust considers such dereferences undefined behavior, and the compiler recently received the facilities to generate checks for this. Thus the unexpected panic message.

Another option would be to substitute u8 for u64:

// trigger a page fault
unsafe {
    *(0xdeadbeef as *mut u8) = 42;
};

[nliao6622]

Yes using the "0xdeadbee0" definitely helps.

[phil-opp]

This should be fixed by #1226.

[SPuntte]

As a meta question, surely the English-language blog post is not supposed to show this "zh-CN" discussion thread?

[phil-opp]

Thanks for reporting this! There was an issue in the comment set up that I just fixed in #1230. I also moved the existing discussions into the correct categories.