linux-kernel - Re: [PATCH v2 8/8] habanalabs: enable 64-bit DMA mask in POWER9

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 for Android: free password hash cracker in your pocket

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

Date:   Wed, 12 Jun 2019 18:17:46 +1000
From:   Benjamin Herrenschmidt <benh@...nel.crashing.org>
To:     "Oliver O'Halloran" <oohall@...il.com>
Cc:     Oded Gabbay <oded.gabbay@...il.com>,
        Greg KH <gregkh@...uxfoundation.org>, linuxppc-dev@...abs.org,
        Christoph Hellwig <hch@...radead.org>,
        Russell Currey <ruscur@....ibm.com>,
        "Linux-Kernel@...r. Kernel. Org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v2 8/8] habanalabs: enable 64-bit DMA mask in POWER9

On Wed, 2019-06-12 at 15:45 +1000, Oliver O'Halloran wrote:
> 
> Also, are you sure about the MSI thing? The IODA3 spec says the only
> important bits for a 64bit MSI are bits 61:60 (to hit the window) and
> the lower bits that determine what IVE to use. Everything in between
> is ignored so ORing in bit 59 shouldn't break anything.

On IODA3... could be different on another system. My point is you can't
just have a fixed setting for all top bits for DMA & MSIs.

> > This will only work as long as all of the system memory can be
> > addressed at an offset from that fixed address that itself fits your
> > device addressing capabilities (50 bits in this case). It may or may
> > not be the case but there's no way to check since the DMA mask logic
> > won't really apply.
> > 
> > You might want to consider fixing your HW in the next iteration... This
> > is going to bite you when x86 increases the max physical memory for
> > example, or on other architectures.
> 
> Yes, do this. The easiest way to avoid this sort of wierd hack is to
> just design the PCIe interface to the spec in the first place.

Ben.