On Mon, Oct 23, 2017 at 11:27 PM, Pavel Machek [off-list ref] wrote:

On Mon 2017-10-23 14:16:40, Linus Walleij wrote:

quoted

On Mon, Oct 23, 2017 at 11:31 AM, Pavel Machek [off-list ref] wrote:

quoted

quoted

quoted

Thinkpad X220... how do I tell if I was using them? I believe so,
because I uncovered bug in them before.

You are certainly using bounce buffers.  What does lspci -knn show?

Here is the output:
0d:00.0 System peripheral [0880]: Ricoh Co Ltd PCIe SDXC/MMC Host Controller [1180:e823] (rev 07)
        Subsystem: Lenovo Device [17aa:21da]
        Kernel driver in use: sdhci-pci

So that is a Ricoh driver, one of the few that was supposed to benefit
from bounce buffers.

Except that if you actually turned it on:

quoted

[10994.302196] kworker/2:1: page allocation failure: order:4,

so it doesn't have enough memory to use these bounce buffers
anyway.

Well, look at archives: driver failed completely when allocation failed.

What I mean is that the allocation probably failed if you
explicitly turned on the bounce buffer also *before*
my patches (like if you were shopping for performance with
the Ricoh driver and turn on bounce buffers) but I haven't tested
it so what do I know.

You could check out b5b6a5f4f06c0624886b2166e2e8580327f0b943
and enable MMC_BLOCK_BOUNCE and see what happens.
And/or benchmark to see if it was actually improving your
system or not.

quoted

I'm now feel it was the right thing to delete them.

Which means I may have been geting benefit -- when it worked. I
believe solution is to allocate at driver probing time.

I think the right way to get this benefit is to enhance the
Ricoh SDMA path with something similar to:
commit 0ccd76d4c236 ("omap_hsmmc: Implement scatter-gather
       emulation")

What it does is loop over the sglist and smatter out one DMA
transfer per sg index.

It's likely faster than copying back and forth to a bounce
buffer even if there is a deal of HW talk back and forth.

(OTOH ... SPI is slow compared to rest of the system, right? Where
does the benefit come from?)

I do not think you will see much performance improvement
on an SPI-based host. Pierre just vaguely remembered "some
Ricoh controllers" would get a benefit from bounce buffers,
no specifics, sorry...

Yours,
Linus Walleij

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>