Fix one more place uio_resid is truncated to int

[cem]

Summary

A follow-up to r231949 and r194990.

[kib]

sys/kern/uipc_mbuf.c

1520 ↗

(On Diff #30109)

This is fine in principle, but qmin() is the wrong function to use. uio_resid is not of quad_t type. Use normal comparision operations there.

[markj]

sys/kern/uipc_mbuf.c

1520 ↗

(On Diff #30109)

Hm, the rest of this function assumes that "total" fits in an int.

[kib]

sys/kern/uipc_mbuf.c

1520 ↗

(On Diff #30109)

It will fit indeed, because len is int. Just that the comparision should use unclipped value of uio_resid.

[cem]

sys/kern/uipc_mbuf.c

1520 ↗

(On Diff #30109)

quad_t encompasses all values ssize_t can have, on current platforms. It's too bad we don't have type-safe min/max macros like Linux.

By normal comparison operations, you mean trinary operator?

[ngie]

sys/kern/uipc_mbuf.c

1520 ↗

(On Diff #30109)

So... uqmin...?

sys/sys/libkern.h:static __inline u_quad_t uqmin(u_quad_t a, u_quad_t b) { return (a < b ? a : b); }

[ngie]

sys/kern/uipc_mbuf.c

1520 ↗

(On Diff #30109)

Actually, szmin?

static __inline ssize_t szmin(ssize_t a, ssize_t b) { return (a < b ? a : b); }

[kib]

sys/kern/uipc_mbuf.c

1520 ↗

(On Diff #30109)

It does not matter what quad_t current domain is.

I mean, use normal C comparision operators like <, be it used in ?: or if is not important. Normal integer promotion rules make it correct.

[cem]

Replace qmin() with expanded form of min.

[ngie]

sys/kern/uipc_mbuf.c

1520 ↗

(On Diff #30136)

Could you please add a comment so someone doesn't reintroduce the bug after they note that the expression is similar to min(..)?

[kib]

sys/kern/uipc_mbuf.c

1520 ↗

(On Diff #30136)

Style states that () should be not used, since they are not needed.

[cem]

sys/kern/uipc_mbuf.c

1520 ↗

(On Diff #30136)

It also makes an allowance for confusion.