netdev - Re: [PATCH] r8169: Fix irq masking in rtl8169

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

Message-Id: <200903200240.50183.fzu@wemgehoertderstaat.de>
Date:	Fri, 20 Mar 2009 02:40:49 +0100
From:	Karsten Wiese <fzu@...gehoertderstaat.de>
To:	Francois Romieu <romieu@...zoreil.com>
Cc:	netdev@...r.kernel.org, Rafael Wysocki <rjw@...k.pl>
Subject: Re: [PATCH] r8169: Fix irq masking in rtl8169_interrupt()

Am Donnerstag 19 März 2009 schrieb Francois Romieu:
> Karsten Wiese <fzu@...gehoertderstaat.de> :
> [...]
> > Found it: CONFIG_DEBUG_SHIRQ is set here causing a "test-call" to
> > rtl8169_interrupt() from irq_request().
> > 
> > $SUBJECT patch is ok, me still thinks. You ?
> 
> I will have to convince myself that nothing bad can happen in the
> window between netif_rx_schedule_prep and RTL_W16. No big deal.

RTL_W16 is a posted write, if it happens, theres no difference if its
posted before or after netif_rx_schedule_prep.

> I do not understand why / how the patch works at all : AFAIU, the status
> register must contain some napi-related event in order for the patch to
> make a difference.

It contains 0x25 when the "test-call" runs, Link, Rx and Tx Flags.

> That's exactly what rtl8169_irq_mask_and_ack() is
> supposed to prevent. -> Does it fail ?

rtl8169_irq_mask_and_ack() resets status once and prevents interrupts
being sent by device. Maybe status reset doesn't work because device is
stopped or device isn't _completely_ stopped and updates status flags later.
Maybe Tx status only becomes 0 if some Tx data is queued.

> So I'd really like to understand what is going on here.

Well, without patch any further interrupt causing napi-status isn't reset
_and_ netif_rx_schedule() isn't active. screaming interrupts.
slooooowish pc. Wonder why it proceeded at all though.

After rmmod/modprobe the status register contained 0x20 when in
rtl8169_irq_mask_and_ack() during "test-call", letting device work
normally.

> PS: while I agree on the motivation for CONFIG_DEBUG_SHIRQ, I wonder if
> it is 100% safe to run an IRQ handler from a context which does not
> adhere to the original semantic. If its {disable/enable}_irq pair races
> with a similar pair due to a different driver on the same (shared) irq,
> things could imho be unsafe.

>From the comment of disable_irq():
	"This function waits for any pending IRQ handlers for this interrupt
 	to complete before returning."
No race possible, in theory ;-)

The pc I debugged this on is a UP with no other device active on the used
int16.

Thanks,
Karsten

--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html