Le 30/07/2021 à 07:10, Greg KH a écrit :
> On Thu, Jul 29, 2021 at 05:32:57PM +0200, Sébastien Rey-Coyrehourcq wrote:
>>   Hi,
>>
>> This is my first bug report on kernel, first crossposted here (https://www.reddit.com/r/NixOS/comments/omctp3/rpi4_nixos_native_2105_and_ssd_by_usb3/)
>> and here on nixos github (https://github.com/NixOS/nixpkgs/issues/130499) :
>>
>> [1.] One line summary of the problem:
>>
>> Using an usb sata adaptater from Startech (USB312SAT3CB 3.1 firmware
>> upgraded) on my RPI4B 8gb with the kernel 5.13.4, during any dd=/ operation
>> on SSD on USB 3 port, my systemd logs finish full of errors of type :
>>
>> juil. 29 14:38:52 Nostromo-sd kernel: xhci_hcd 0000:01:00.0: swiotlb buffer
>> is full (sz: 512 bytes), total 32768 (slots), used 32768 (slots)
>>
>> and
>>
>> juil. 29 09:25:07 Nostromo-sd kernel: ------------[ cut here ]------------
>> juil. 29 09:25:07 Nostromo-sd kernel: WARNING: CPU: 1 PID: 180 at
>> drivers/mmc/host/sdhci.c:1142 sdhci_prepare_data.isra.0+0x298/0x700
> Your problem is in the sdhci code, not the usb controller.
>
> But even with this, does the operation succeed properly?

Nop, operation failed, my systemd journal is full of these errors, 4gb 
at the end,
and finally the dd command and system freeze at less than 15% of the dd 
operation.

> The rpi hardware is HORRID when it comes to the USB controller, so those
> xhci messages might just be normal given the lack of irq handling that
> the platform was known to have in the past (note, might be better now,
> no idea, sorry.)
>
> But again, is the operation failing?  Is your sd card working properly?
> Does the problem go away with a new/better sd card?

The sd is fresh new but, i could will try with another if the problem is 
linked to sd.
My final objective was to use boot on SSD feature and not a slow SD for 
my NIxos system support on RPI, so i suppose i could put this problem 
aside, and prepare my SSD with another computer..

Thanks,

> thanks,
>
> greg k-h