[I guess there is nothing secret in this mail and moreover it touches more people than me. Thus, Cc'ed to dmaengine@ mailing list. In the future please avoid private-only messages on such topics.] On Wed, 2015-09-09 at 17:26 +0200, Vitaly Bordyug wrote: > Hello Andy, > > sorry for reaching out directly, but the question is very dmatest > -specific anyways. > > As far as I understood from the code, dmatest engages mem-to-mem > transactions on dma cannel awailable, using basically small chunks > allocated in kernel memory as a payload. > > What I need at the moment, is an analog of mtest which uses as much > dma as possible (that is to say, stressing SDRAM (and dmaengine) not > involving much of the CPU) . > > Hence, I am now thinking on limiting the kernel memory with mem=X, > ioremapping the top part, and extending the dmatest to work with raw > offsets and significantly bigger data chunks. > > Is it doable or am I suggesting something completely stupid? > > Sorry again about direct bothering and thanks in advance! > > Sincerely, > > -Vitaly -- Andy Shevchenko <andriy.shevchenko@xxxxxxxxxxxxxxx> Intel Finland Oy -- To unsubscribe from this list: send the line "unsubscribe dmaengine" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html