renesas/vsp-tests.git - VSP test framework

Age	Commit message (Collapse)	Author
2017-02-13	tests: Add basic stress testing	Kieran Bingham
	Duplicate the functionality of vsp-unit-test-0001, to provide a set of work for the hardware to run, whilst we load the system using the 'stress' utility. Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com> Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>

09:03 < wsa> welcome everyone to todays meeting 09:03 < wsa> here are the collected status updates 09:03 < wsa> Status updates 09:03 < wsa> ============== 09:03 < wsa> A - what have I done since last time 09:03 < wsa> ------------------------------------ 09:03 < wsa> Geert 09:03 < wsa> : implemented explicit internal delay setting from DT for ravb, DT binding doc conversion to json-schema: ravb 09:03 < wsa> Shimoda-san 09:03 < wsa> : converted XHCI binding docs to YAML, fixed undefined temperature if negative for Thermal, and sent various series to fix mmc_suspend() with PSCI 09:03 < wsa> Ulrich 09:03 < wsa> : sent new series to handle WDT handover from bootloader which got merged, sent new series implementing atomic transfers for IIC 09:03 < wsa> Wolfram 09:03 < wsa> : removed final bits of i2c_new_device API and removed the old API so API conversion is now finished, reviewed slave implementation for SMBusHostNotify, and added the support for i2c-rcar, fixed various corner cases in the slave implemetation of i2c-rcar, wrote a new slave-backend to test rare I2C/SMBus features which usually need specific devices, tried to get an answer to the RPM 09:03 < wsa> refcounting issue, reviewed Uli's atomic transfer patch for IIC, minor treewide cleanup patches sent out while working on all of the above 09:03 < wsa> B - what I want to do until next time 09:03 < wsa> ------------------------------------- 09:03 < wsa> Geert 09:03 < wsa> : wants to implement generic bindings for Ethernet MAC explicit internal delay setting 09:03 < wsa> Niklas 09:03 < wsa> : wants to eet up with Wolfram and flash Gen3 boards, and do the yearly test with SDIO-WIFI on Koelsch 09:03 < wsa> Shimoda-san 09:03 < wsa> : wants to keep working on mmc_suspend() with PSCI, discuss with Cogent about RAVB driver patch, convert usb-xhci, rcar-pci and sdhi DT docs to json-schema 09:03 < wsa> Ulrich 09:03 < wsa> : wants to sent new series implementing atomic transfers for IIC and I2C 09:03 < wsa> Wolfram 09:03 < wsa> : wants to meet with Niklas and fix Gen3 boards together, have two weeks of holidays, add SMBusAlert and I2C_M_RECV_LEN to both, the slave-testunit and the i2c-rcar driver, resend series 'fix stalled SCC' and 'implement manual calibration' for SDHI, review Shimoda-san's DMA unmapping series for SDHI 09:03 < wsa> C - problems I currently have 09:03 < wsa> ----------------------------- 09:03 < wsa> Shimoda-san 09:03 < wsa> : wonders whether the mmc_suspend() improvement is really needed because we cannot avoid accidentally power off of eMMC around system suspend anyway 09:03 < wsa> Wolfram 09:03 < wsa> : asks how to ensure that atomic xfers on IIC have their clock enabled that late, and couldn't use Gen3 boards for a while so he had to switch to work which could be tested on Gen2 09:04 < wsa> shimoda: I assume with the RAVB patch you mentioned is about the "stop queues on timeout" issue? 09:05 < wsa> geertu: you mentioned that you'd apply Uli's RWDT clock patches but I don't see them in -next yet 09:05 < shimoda> wsa: yes 09:05 < geertu> wsa: They're in clk-renesas. Haven't sent a pull request to mturquette yet 09:06 < wsa> geertu: ah, they don't go directly into next from your branch? 09:06 * geertu doublechecks 09:06 < geertu> wsa: No they don't. Same with sh-pfc. 09:06 < wsa> shimoda: ok, thanks. will update that information 09:06 < shimoda> wsa: thanks! 09:07 < wsa> about the RPM get_sync ref-counting issue: 09:07 < geertu> shimoda: I assume s/Cogent/Sergei/? 09:07 < shimoda> geertu: yes 09:08 < wsa> I confirmed that always increasing the ref-count is intended behaviour. I have not confirmed yet, that it may be a better fix to simply remove the error checking instead of adding a put_noidle everywhere 09:08 < wsa> but from the last discussion, I thought for R-Car this is true? 09:09 < geertu> wsa: Yes, Rafael confirmed not checking is fine for us 09:09 < wsa> ok, nice to know 09:10 < wsa> so, i will ask people sending in such patches if they checked that removing the check may be the better solution 09:10 < geertu> https://lore.kernel.org/r/CAJZ5v0i87NGcy9+kxubScdPDyByr8ypQWcGgBFn+V-wDd69BHQ@mail.gmail.com 09:11 < geertu> "It is the right thing to do in the majority of cases." 09:11 < geertu> Now, there's still the bunch of "fixers" that will try to add error paths (and pm_runtime_put_*() calls) everywhere 09:12 < wsa> maybe we should add another function annotation: __must_not_check ;) 09:13 < wsa> at least, I now have a proper answer for them 09:13 < geertu> ;-) 09:14 < wsa> so, I have another topic but wanted to ask you guys first if there are questions regarding the status updates or in general? 09:14 < neg> wsa: is not __must_not_check equal to returning void :-) 09:15 < wsa> shimoda: I will seriously try to review the DMA unbalance patches before my holidays 09:15 < geertu> wsa: For atomic transfers, it's not just the clock, but also the power domain 09:15 < wsa> neg: you got it! 09:16 < wsa> geertu: yes, this was my remaining question 09:17 < shimoda> wsa: thanks! 09:17 < geertu> It seems some modules on R-Car Gen2 are "forgiving" when accessing them with the module clock disabled (i2c, rwdt) 09:17 < wsa> atomic_xfer happen very late, so clocks and power domains may be off already. and we can't use the calls to enable them anymore, right? 09:17 < geertu> Yes, pm_runtime_get_sync() may sleep 09:18 < geertu> We might set GENPD_FLAG_IRQ_SAFE for our domains 09:19 < wsa> but on what condition? 09:19 * kbingham discovers http://sweng.the-davies.net/Home/rustys-api-design-manifesto :-) 09:19 < geertu> kbingham: Yeah, pm_runtime_get_sync() is at -1 ;-) 09:19 < kbingham> geertu, That's the thread I found it from ;-) 09:20 < pinchartl> did Rafael provide any answer regarding whether pm_runtime_get_sync() should be fixed ? 09:20 < wsa> I thought it was -4 09:20 < geertu> there's also dev.power.irq_safe