usb-serial: optimize sysrq function calls
in [Kernel]
|
Prev: [PATCH] MAINTAINERS: add bfin_sdh driver
Next: [PATCH -tip][URGENT] kprobes: Calculate the index correctly when freeing the out-of-line execution slot.
From: Alan Stern on 9 Mar 2010 10:20 On Tue, 9 Mar 2010, Jason Wessel wrote: > There is no need to have external function calls for the sysrq > functions. The compiler can inline the sysrq calls such that they are > entirely a NOP if CONFIG_MAGIC_SYSRQ is not set. This is not the best way to do what you want. Keep usb_serial_handle_sysrq_char() and usb_serial_handle_break() as out-of-line routines in generic.c, but make them conditional on CONFIG_MAGIC_SYSRQ. Then in the header file, depending on whether or not CONFIG_MAGIC_SYSRQ is defined, either put the usual extern function declarations or else put do-nothing inline definitions. The advantage of keeping the functions out-of-line is the reduced amount of code space (since the bodies aren't replicated every place they get used). This way you keep that advantage while still getting the NOP implementation if CONFIG_MAGIC_SYSRQ isn't set. Alan Stern -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo(a)vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
From: Jason Wessel on 10 Mar 2010 17:20
Alan Stern wrote: > On Tue, 9 Mar 2010, Jason Wessel wrote: > > >> There is no need to have external function calls for the sysrq >> functions. The compiler can inline the sysrq calls such that they are >> entirely a NOP if CONFIG_MAGIC_SYSRQ is not set. >> > > This is not the best way to do what you want. Keep > usb_serial_handle_sysrq_char() and usb_serial_handle_break() as > out-of-line routines in generic.c, but make them conditional on > CONFIG_MAGIC_SYSRQ. Then in the header file, depending on whether or > not CONFIG_MAGIC_SYSRQ is defined, either put the usual extern function > declarations or else put do-nothing inline definitions. > > The advantage of keeping the functions out-of-line is the reduced > amount of code space (since the bodies aren't replicated every place > they get used). This way you keep that advantage while still getting > the NOP implementation if CONFIG_MAGIC_SYSRQ isn't set. > It is probably fine to just drop this patch entirely then. I had originally made the change to keep it similar to the uart based serial drivers where the definitions are inlined. Jason. -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo(a)vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/ |