rpma: Atomic write with IBV_SEND_FENCE flag

[ewanglong]

This change is 

[janekmi]

Reviewed 1 of 1 files at r1.
Reviewable status: all files reviewed, 1 unresolved discussion (waiting on @ewanglong)

---

src/mr.c, line 171 at r1 (raw file):

	wr.send_flags = (flags & RPMA_F_COMPLETION_ON_SUCCESS) ?
		IBV_SEND_SIGNALED : 0;
	wr.send_flags |= (flags & RPMA_F_COMPLETION_ON_ERROR) ?

I do not understand the condition here. Can you elaborate?

[grom72]

Reviewable status: 0 of 9 files reviewed, 2 unresolved discussions (waiting on @ewanglong and @janekmi)

---

src/mr.h, line 41 at r2 (raw file):

int fence

why is this not combined with other flags?

[ldorau]

Reviewed 3 of 9 files at r2.
Reviewable status: 3 of 9 files reviewed, 2 unresolved discussions (waiting on @ewanglong)

[grom72]

Reviewed 1 of 1 files at r1, 8 of 9 files at r2.
Reviewable status: all files reviewed, 7 unresolved discussions (waiting on @ewanglong)

---

src/mr.h, line 41 at r2 (raw file):

int

bool is enough

---

src/mr.c, line 171 at r1 (raw file):

flags & RPMA_F_COMPLETION_ON_ERROR

fence & RPMA_F_SEND_FENCE

---

src/include/librpma.h, line 426 at r2 (raw file):

#define RPMA_F_SEND_NOFENCE		0
#define RPMA_F_SEND_FENCE		1

Shall it be in mr.h? as it is not a part of public API yet

---

tests/unit/common/test-common.h, line 33 at r2 (raw file):

#define MOCK_NOFENCE		0
#define MOCK_FENCE			1

shall we use RPMA_F_SEND_NOFENCE and RPMA_F_SEND_FENCE instead of MOCK_(NO)FENCE
or just True/False if we use bool instead of int?

---

tests/unit/mr/mr-write.c, line 23 at r2 (raw file):

 * with RPMA_E_PROVIDER when send_flags == RPMA_F_COMPLETION_ON_SUCCESS
 */
static void

tests for FENCE are missing

[ldorau]

Reviewable status: 3 of 8 files reviewed, 7 unresolved discussions (waiting on @ewanglong, @grom72, @janekmi, and @ldorau)

---

src/mr.h, line 41 at r2 (raw file):

Previously, grom72 (Tomasz Gromadzki) wrote…

int fence

why is this not combined with other flags?

Why is bool fence not merged with int flags?

---

src/mr.c, line 171 at r1 (raw file):

Previously, janekmi (Jan Michalski) wrote…

I do not understand the condition here. Can you elaborate?

I do not understand it either, wr.send_flags can have only IBV_* values

[ldorau]

Reviewed 4 of 6 files at r3.
Reviewable status: 6 of 8 files reviewed, 7 unresolved discussions (waiting on @ewanglong, @grom72, and @janekmi)

[grom72]

Reviewed 6 of 6 files at r3.
Reviewable status: all files reviewed, 6 unresolved discussions (waiting on @ewanglong, @janekmi, and @ldorau)

---

src/mr.h, line 41 at r2 (raw file):

Previously, ldorau (Lukasz Dorau) wrote…

Why is bool fence not merged with int flags?

because flags is related to public API while fence is an internal feature

[janekmi]

Reviewed 3 of 9 files at r2, 5 of 6 files at r3.
Reviewable status: all files reviewed, 6 unresolved discussions (waiting on @ewanglong, @grom72, and @ldorau)

---

src/mr.c, line 172 at r3 (raw file):

		IBV_SEND_SIGNALED : 0;
	wr.send_flags |= (fence == true) ?
		IBV_SEND_FENCE : 0;

I think this can fit into the previous line.

---

tests/unit/common/test-common.h, line 33 at r2 (raw file):

Previously, grom72 (Tomasz Gromadzki) wrote…

#define MOCK_NOFENCE		0
#define MOCK_FENCE			1

shall we use RPMA_F_SEND_NOFENCE and RPMA_F_SEND_FENCE instead of MOCK_(NO)FENCE
or just True/False if we use bool instead of int?

I like it the way it is.

[janekmi]

Reviewable status: all files reviewed, 7 unresolved discussions (waiting on @ewanglong, @grom72, and @ldorau)

a discussion (no related file):
Please use common: prefix for your commit.

---

[grom72]

Reviewed 1 of 1 files at r4.
Reviewable status: all files reviewed, 3 unresolved discussions (waiting on @ewanglong)

[janekmi]

Reviewed 1 of 1 files at r4.
Reviewable status: all files reviewed, 1 unresolved discussion (waiting on @ewanglong)

[grom72]

Reviewable status: all files reviewed, 2 unresolved discussions (waiting on @ewanglong)

---

src/mr.c, line 171 at r3 (raw file):

fence == true

fence instead of fence == true

and the rest of operator ? in the same line

[grom72]

Reviewable status: 7 of 8 files reviewed, 1 unresolved discussion (waiting on @ewanglong, @grom72, @janekmi, and @ldorau)

[grom72]

Reviewed 1 of 1 files at r5.
Reviewable status: all files reviewed, 1 unresolved discussion (waiting on @ewanglong)

[ldorau]

Reviewed 2 of 9 files at r2, 2 of 6 files at r3, 1 of 1 files at r5.
Reviewable status: complete! all files reviewed, all discussions resolved (waiting on @ewanglong)

[ldorau]

Reviewed 1 of 1 files at r4.
Reviewable status: complete! all files reviewed, all discussions resolved (waiting on @ewanglong)