> > > On Thu, 21 Feb 2019 at 08:16, hongjiefang <hongjiefang@xxxxxxxxxxxx> > wrote: > > > > > > > > when failed to verify the cid, the card initialisation will abort. > > > > it is necessary to print some logs for this critical error. > > > > > > Under what circumstances do you observe this error? Is it during > > > system resume or during runtime resume? > > > > > > Moreover, to get this error, I guess you are remove one card and > > > replacing it with another, right? > > > > It happened when the system resume a non-removable eMMC card > > from deep suspend. > > > > > Just wanted to be sure of what kind of scenario you are > > > testing/looking at. Perhaps you could even fold in some of that > > > information in the changelog, that would be nice. > > > > During the eMMC resume stage, CMD2 returned an error CID value > > but without any error INT status. > > The error CID values are as follow 15010044, 4436384d, 4202c2c0, ff808000. > > The correct value should be 15010044, 4436384d, 4202c2d8, dc2f9400. > > "c0ff8080" is the response from CMD1. > > I see, that is indeed a weird error case. There is for sure a more > serious bug somewhere to look for. > > I guess a overwriting the memory containing the response data, > somewhere from the host driver up until here is the is what one should > be looking for in the at first place. > > Of course, another option is that the eMMC is internally behaving > in-correct, but then you need to have the debug equipment to be able > to listen and decode the bits transferred on the physical lines. > > > > > This problem occasionally occurred when system resume from suspend status. > > Maybe it happened because there is something wrong about bus clock or our > host > > when system do resume. It needs further debugging. > > I see, but in such case, there should be an error printed in > mmc_bus_resume(), along the lines of "...card was removed?"... > > Isn't that sufficient? There is an error log " mmc0: error -2 doing runtime resume". But there's no clear indication of where went wrong. This patch wants to give a clear hint. It should be better. > > [...] > > Kind regards > Uffe B&R Hongjie
