usb(4): Stop checking for failures from malloc(M_WAITOK).

[markj]

Summary

PR: 240545
MFC after: 1 week

[emaste]

LGTM

[• hselasky]

The USB driver is also used in boot loader environments where malloc may fail even if M_WAITOK is set. Maybe M_WAITOK should be conditionally set?

[• hselasky]

I've marked the two cases which can happen from bootloader.

Maybe you can do something like this:

#define USB_HAVE_MALLOC_WAITOK 0

#define USB_HAVE_MALLOC_WAITOK 1

In:

sys/dev/usb/usb_freebsd_loader.h:#define USB_HAVE_UGEN 0
sys/dev/usb/usb_freebsd.h:#define USB_HAVE_UGEN 1

sys/dev/usb/usb_device.c

1788 ↗

(On Diff #74573)

This one can happen in bootloader.

sys/dev/usb/usb_transfer.c

1346 ↗

(On Diff #74573)

This one can happen in bootloader.

[markj]

• Add USB_HAVE_MALLOC_WAITOK, use it in two places.
• Remove more M_WAITOK handling in ugen_ioctl_post().

[markj]

In D25706#569107, @hselasky wrote:

I've marked the two cases which can happen from bootloader.

Maybe you can do something like this:

#define USB_HAVE_MALLOC_WAITOK 0

#define USB_HAVE_MALLOC_WAITOK 1

In:

sys/dev/usb/usb_freebsd_loader.h:#define USB_HAVE_UGEN 0
sys/dev/usb/usb_freebsd.h:#define USB_HAVE_UGEN 1

Hmm, thanks. I added some checking for the cases you pointed out. It might make sense to have a usb_malloc() that can take three flags: USB_NOWAIT (always can fail), USB_WAITOK (always expected to succeed), USB_TRYWAIT (use M_WAITOK if the platform guarantees that it will succeed, else use M_NOWAIT).

[• hselasky]

In D25706#569755, @markj wrote:

Hmm, thanks. I added some checking for the cases you pointed out. It might make sense to have a usb_malloc() that can take three flags: USB_NOWAIT (always can fail), USB_WAITOK (always expected to succeed), USB_TRYWAIT (use M_WAITOK if the platform guarantees that it will succeed, else use M_NOWAIT).

I like the current approach too ...