linux-kernel - Re: [PATCH kvm RESEND] KVM: i8259: Fix poll command

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Date:   Mon, 10 Apr 2023 09:32:15 -0700
From:   Sean Christopherson <seanjc@...gle.com>
To:     alexjlzheng@...il.com
Cc:     pbonzini@...hat.com, tglx@...utronix.de, mingo@...hat.com,
        bp@...en8.de, dave.hansen@...ux.intel.com, x86@...nel.org,
        hpa@...or.com, kvm@...r.kernel.org, linux-kernel@...r.kernel.org,
        Jinliang Zheng <alexjlzheng@...cent.com>
Subject: Re: [PATCH kvm RESEND] KVM: i8259: Fix poll command

Please don't use RESEND as a ping, just respond to the original patch with a "Ping",
or any question you might have.  I know Documentation/process/submitting-patches.rst
says its ok to RESEND after a couple of weeks, but IMO that's overly aggressive
and just creates noise, e.g. your original patch was in my todo list, I just hadn't
gotten too it.  If you can't get a response after multiple pings, then by all means
RESEND, but in the future, please try pinging first.

For the patch context, there's no need to put "kvm" after patch, i.e. [PATCH], or
in this case [PATCH RESEND].  The "KVM:" namespace in the shortlog provides
sufficient context.

Regarding the shortlog, if a v2 is needed, ignore the somewhat messy history of
this file and use "KVM: x86:".

On Mon, Apr 10, 2023, alexjlzheng@...il.com wrote:
> From: Jinliang Zheng <alexjlzheng@...cent.com>
> 
> According to the hardware manual, when the Poll command is issued, the
> byte returned by the I/O read is 1 in Bit 7 when there is an interrupt,
> and the highest priority binary code in Bits 2:0. The current pic
> simulation code is not implemented strictly according to the above
> expression.

There is way too much going on in this patch for this to be a sufficient description.
pic_intack() is not a direct replacement for the open coded logic in pic_poll_read(),
modulo the setting of bit 7.  E.g. there's no explanation for the "addr1 >> 7"
logic, pic_clear_isr() is conditionally called on auto_eoi, priority_add is now
modified, pic_update_irq() is no longer called, and so on and so forth.

Maybe the patch is correct and pic_poll_read() was completely broken, but if that's
the case, the changelog needs to be _much_ more verbose in explaining everything.

> Fix the implementation of poll mode in pic simulation by pic_intack,

Add () when referencing functions by name, i.e. pic_intack().

> and remove redundant pic_poll_read code.

Removing pic_poll() needs to be done in a separate patch.  Removing the helper
while simultaneously modifying its effective code makes the patch unnecessarily
difficult to review.

> Signed-off-by: Jinliang Zheng <alexjlzheng@...cent.com>
> ---
>  arch/x86/kvm/i8259.c | 29 ++++++-----------------------
>  1 file changed, 6 insertions(+), 23 deletions(-)
> 
> diff --git a/arch/x86/kvm/i8259.c b/arch/x86/kvm/i8259.c
> index 4756bcb5724f..bc5b758e8f73 100644
> --- a/arch/x86/kvm/i8259.c
> +++ b/arch/x86/kvm/i8259.c
> @@ -397,35 +397,18 @@ static void pic_ioport_write(void *opaque, u32 addr, u32 val)
>  		}
>  }
>  
> -static u32 pic_poll_read(struct kvm_kpic_state *s, u32 addr1)
> -{
> -	int ret;
> -
> -	ret = pic_get_irq(s);
> -	if (ret >= 0) {
> -		if (addr1 >> 7) {
> -			s->pics_state->pics[0].isr &= ~(1 << 2);
> -			s->pics_state->pics[0].irr &= ~(1 << 2);
> -		}
> -		s->irr &= ~(1 << ret);
> -		pic_clear_isr(s, ret);
> -		if (addr1 >> 7 || ret != 2)
> -			pic_update_irq(s->pics_state);
> -	} else {
> -		ret = 0x07;
> -		pic_update_irq(s->pics_state);
> -	}
> -
> -	return ret;
> -}
> -
>  static u32 pic_ioport_read(void *opaque, u32 addr)
>  {
>  	struct kvm_kpic_state *s = opaque;
>  	int ret;
>  
>  	if (s->poll) {
> -		ret = pic_poll_read(s, addr);
> +		ret = pic_get_irq(s);
> +		if (ret >= 0) {
> +			pic_intack(s, ret);
> +			ret |= 0x80;
> +		} else

All branches in an if-elif-else statment need curly braces if any branch needs
statements (again, ignore the bad "prior art" in this file), i.e.

		if (ret >= 0) {
			...
		} else {
			ret = 0;
		}

> +			ret = 0;
>  		s->poll = 0;
>  	} else
>  		if ((addr & 1) == 0)
> -- 
> 2.37.3
>