On Tue, 25 Feb 2020, Eugeniu Rosca wrote: > Hi Alan, > > Many thanks for the prompt review. > > On Tue, Feb 25, 2020 at 10:32:32AM -0500, Alan Stern wrote: > > On Tue, 25 Feb 2020, Eugeniu Rosca wrote: > > > + int r = usb_autopm_get_interface(intf); > > > + > > > + if (!r) > > > + hub->quirk_disable_autosuspend = 1; > > > + else > > > + dev_dbg(&intf->dev, "disable autosuspend err=%d\n", r); > > > } > > > > > > if (hub_configure(hub, &desc->endpoint[0].desc) >= 0) > > > > This change is not necessary, because the resume operation cannot fail > > at this point (interfaces are always powered-up during probe). > > Agreed to avoid unneeded complexity. > > > A better solution would be to call usb_autpm_get_interface_no_resume() > > instead. > > Pushed to https://lore.kernel.org/lkml/20200225183057.31953-1-erosca@xxxxxxxxxxxxxx > > > > > On the other hand, the other places that call > > usb_autopm_get_interface() without checking should be audited. Some of > > them almost certainly need to be fixed. > > A quick 'git grep' outputs below list of auditable candidates [1]. > > With no relevant devices at hand, I would avoid touching these drivers, > since oftentimes even legitimate patches introduce regressions w/o > testing. > > If anybody volunteers with testing, I guess it should be quick to > either convert usb_autpm_get_interface to *_no_resume variant or > handle the return value in place in below instances. > > [1] (v5.6-rc3) git grep -En "[^=]\s+usb_autopm_get_interface\(" > drivers/input/touchscreen/usbtouchscreen.c:1788: usb_autopm_get_interface(intf); > drivers/media/usb/stkwebcam/stk-webcam.c:628: usb_autopm_get_interface(dev->interface); > drivers/net/usb/hso.c:1308: usb_autopm_get_interface(serial->parent->interface); > drivers/net/usb/r8152.c:5231: usb_autopm_get_interface(tp->intf); > sound/usb/usx2y/us122l.c:192: usb_autopm_get_interface(iface); Your regular expression isn't right because it doesn't match lines where the usb_autopm_get_interface() is preceded only by one whitespace character (i.e., a tab). It also will match lines where there are two space characters between the = sign and the function name. I think what you want is more like "(^|[^=[:space:]])\s*" at the start of the RE. A revised search finds line 997 in drivers/usb/core/hub.c and lines 216, 269 in drivers/usb/core/port.c. (I didn't try looking in any other directories.) AFAICT all three of these should check the return value, although a error message in the kernel log probably isn't needed. Do you want to fix those instances up also, maybe merging them in with the existing patch? Alan Stern
