linux-kernel - Re: [PATCH v6 0/2] PCI: dwc: Add support for 64-bit MSI target addresses

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]

Date:   Tue, 31 Jan 2023 12:42:57 +0000
From:   Robin Murphy <robin.murphy@....com>
To:     Evgenii Shatokhin <e.shatokhin@...ro.com>,
        Will McVicker <willmcvicker@...gle.com>,
        Lorenzo Pieralisi <lpieralisi@...nel.org>
Cc:     kernel-team@...roid.com, Vidya Sagar <vidyas@...dia.com>,
        Christoph Hellwig <hch@...radead.org>,
        linux-pci@...r.kernel.org, linux-kernel@...r.kernel.org,
        Gustavo Pimentel <gustavo.pimentel@...opsys.com>,
        Jingoo Han <jingoohan1@...il.com>,
        Rob Herring <robh@...nel.org>,
        Krzysztof Wilczyński <kw@...ux.com>,
        Bjorn Helgaas <bhelgaas@...gle.com>, linux@...ro.com
Subject: Re: [PATCH v6 0/2] PCI: dwc: Add support for 64-bit MSI target
 addresses

On 2023-01-31 12:29, Evgenii Shatokhin wrote:
> Hi,
> 
> On 26.08.2022 02:54, Will McVicker wrote:
>> Hi All,
>>
>> I've update patch 2/2 to address Robin's suggestions. This includes:
>>
>>   * Dropping the while-loop for retrying with a 64-bit mask in favor of
>>     retrying within the error if-statement.
>>   * Using an int for the DMA mask instead of a bool and ternary 
>> operation.
>>
>> Thanks again for the reviews and sorry for the extra revision today!
>> Hopefully this is the last one :) If not, I'd be fine to submit patch 1/2
>> without 2/2 to avoid resending patch 1/2 for future revisions of patch 
>> 2/2
>> (unless I don't need to do that anyway).
> 
> The first patch of the series made it into the mainline kernel, but, it 
> seems, the second one ("PCI: dwc: Add support for 64-bit MSI target 
> address") did not. As of 6.2-rc6, it is still missing.
> 
> Was it intentionally dropped because of some issues or, perhaps, just by 
> accident? If it was by accident, could you please queue it for inclusion 
> into mainline again?

Yes, it was dropped due to the PCI_MSI_FLAGS_64BIT usage apparently 
being incorrect, and some other open debate (which all happened on the 
v5 thread):

https://lore.kernel.org/linux-pci/YzVTmy9MWh+AjshC@lpieralisi/

The DMA mask issues have now been sorted out, so you, or Will, or anyone 
else interested should be free to rework this on top of linux-next 
(although at this point, more realistically on top of 6.3-rc1 in a few 
weeks).

Thanks,
Robin.

> Support for 64-bit MSI target addresses is needed for some of our SoCs. 
> I ran into a situation when there was no available RAM in ZONE_DMA32 
> during initialization of PCIe host. Hence, dmam_alloc_coherent() failed 
> in dw_pcie_msi_host_init() and initialization failed with -ENOMEM:
> 
> [    0.374834] dw-pcie 4000000.pcie0: host bridge /soc/pcie0@...0000 
> ranges:
> [    0.375813] dw-pcie 4000000.pcie0:      MEM 
> 0x0041000000..0x004fffffff -> 0x0041000000
> [    0.376171] dw-pcie 4000000.pcie0:   IB MEM 
> 0x0400000000..0x07ffffffff -> 0x0400000000
> [    0.377914] dw-pcie 4000000.pcie0: Failed to alloc and map MSI data
> [    0.378191] dw-pcie 4000000.pcie0: Failed to initialize host
> [    0.378255] dw-pcie: probe of 4000000.pcie0 failed with error -12
> 
> Mainline kernel 6.2-rc6 was used in that test.
> 
> The hardware supports 64-bit target addresses, so the patch "PCI: dwc: 
> Add support for 64-bit MSI target address" should help with this 
> particular failure.
> 
> 
>>
>> Thanks,
>> Will
>>
>> Will McVicker (2):
>>    PCI: dwc: Drop dependency on ZONE_DMA32
>>
>> v6:
>>   * Retrying DMA allocation with 64-bit mask within the error 
>> if-statement.
>>   * Use an int for the DMA mask instead of a bool and ternary operation.
>>
>> v5:
>>   * Updated patch 2/2 to first try with a 32-bit DMA mask. On failure,
>>     retry with a 64-bit mask if supported.
>>
>> v4:
>>   * Updated commit descriptions.
>>   * Renamed msi_64b -> msi_64bit.
>>   * Dropped msi_64bit ternary use.
>>   * Dropped export of dw_pcie_msi_capabilities.
>>
>> v3:
>>    * Switched to a managed DMA allocation.
>>    * Simplified the DMA allocation cleanup.
>>    * Dropped msi_page from struct dw_pcie_rp.
>>    * Allocating a u64 instead of a full page.
>>
>> v2:
>>    * Fixed build error caught by kernel test robot
>>    * Fixed error handling reported by Isaac Manjarres
>>   PCI: dwc: Add support for 64-bit MSI target address
>>
>>   .../pci/controller/dwc/pcie-designware-host.c | 43 +++++++++----------
>>   drivers/pci/controller/dwc/pcie-designware.c  |  8 ++++
>>   drivers/pci/controller/dwc/pcie-designware.h  |  2 +-
>>   3 files changed, 30 insertions(+), 23 deletions(-)
>>
>>
>> base-commit: 568035b01cfb107af8d2e4bd2fb9aea22cf5b868
> 
> Thank you in advance.
> 
> Regards,
> Evgenii
> 
> 
>