On 19 May 2016 at 01:52, Peter Chen <hzpeterchen@xxxxxxxxx> wrote: > > > On Thu, May 19, 2016 at 5:24 AM, Arnd Bergmann <arnd@xxxxxxxx> wrote: >> >> I stumbled over this warning last week, which showed up after I had >> removed an incorrect patch from my randconfig build setup: >> >> drivers/usb/phy/phy-msm-usb.c: In function 'msm_otg_probe': >> drivers/usb/phy/phy-msm-usb.c:1735:14: error: 'regs[0].consumer' may be >> used uninitialized in this function [-Werror=maybe-uninitialized] >> motg->vddcx = regs[0].consumer; >> ~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~ >> drivers/usb/phy/phy-msm-usb.c:1736:14: error: 'regs[1].consumer' may be >> used uninitialized in this function [-Werror=maybe-uninitialized] >> motg->v3p3 = regs[1].consumer; >> ~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~ >> drivers/usb/phy/phy-msm-usb.c:1737:14: error: 'regs[2].consumer' may be >> used uninitialized in this function [-Werror=maybe-uninitialized] >> motg->v1p8 = regs[2].consumer; >> ~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~ >> >> Having already fixed the same problem in the phy-qcom-8x16-usb.c >> driver before, I tried to do the same thing here, but it turned >> out to be somewhat different, and I ended up running into several >> unrelated issues in the driver that I now try to fix up. >> >> The series is not tested beyond verifying that it no longer causes >> randconfig warnings, and some patches are not entirely obvious. >> In particular the ehci-msm and chipidea changes are probably a >> good idea, but they actually change the behavior of the drivers >> in a way that I cannot verify through inspection alone. The >> last patch in the series probably requires some changes to the >> devicetree files to go along with it. >> >> Please have a look and test this, provided the patches make sense >> conceptually. >> >> > > > MSM USB driver uses phy driver for dual-role (otg) switch, chipidea > driver for peripheral function, a dedicated ehci file for host driver, > so these three driver will access the same controller register space, > it is not a good practice. Agreed. It's a huge mess. And it is currently being reworked. > In fact, MSM USB2 can use chipidea driver as controller driver, and its > phy driver dedicate for USB PHY function with some changes. Yeah I noticed with the addition of the host controller we can do this, which was the impetus for reworking. In the end, we'll have just the qcom-ci and a phy. Regards, Andy -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
