rpma: adopt usage_to_access to support iWARP specificity

[ldorau]

This change is 

[codecov]

Codecov Report

Merging #1044 (e442b98) into master (100789d) will increase coverage by 0.00%.
The diff coverage is 100.00%.

❗ Current head e442b98 differs from pull request most recent head c8c0323. Consider uploading reports for the commit c8c0323 to get more accurate results

@@           Coverage Diff           @@
##           master    #1044   +/-   ##
=======================================
  Coverage   99.83%   99.84%           
=======================================
  Files          15       15           
  Lines        1247     1251    +4     
=======================================
+ Hits         1245     1249    +4     
  Misses          2        2           

[janekmi]

Reviewed 10 of 10 files at r1.
Reviewable status: all files reviewed, 4 unresolved discussions (waiting on @ldorau)

---

src/peer.c, line 32 at r1 (raw file):

};

/* helper functions */

This is not a helper any more. It is an internal API now.

---

src/peer.c, line 54 at r1 (raw file):

		access |= IBV_ACCESS_LOCAL_WRITE;

		/* XXX */

Just a reminder.

---

tests/unit/peer/peer-common.h, line 28 at r1 (raw file):

 * RPMA_MR_USAGE_* values.
 */
#define MOCK_USAGE		(unsigned)255

Can we just make it a combination instead of using a hardcoded value?

---

tests/unit/peer/peer-mr_reg.c, line 40 at r1 (raw file):

	struct ibv_mr *mr = NULL;
	int ret = rpma_peer_mr_reg(peer, &mr, MOCK_ADDR,
				MOCK_LEN, MOCK_USAGE);

Since the usage -> access calculation is now a part of the peer module I think also the testing of it should go here.
Moreover, we will need two protection domains: one iWARP and one non-iWARP to test both implementations.

[ldorau]

Reviewable status: 8 of 10 files reviewed, 4 unresolved discussions (waiting on @janekmi)

---

src/peer.c, line 32 at r1 (raw file):

Previously, janekmi (Jan Michalski) wrote…

This is not a helper any more. It is an internal API now.

Done.

---

src/peer.c, line 54 at r1 (raw file):

Previously, janekmi (Jan Michalski) wrote…

Just a reminder.

Done.

---

tests/unit/peer/peer-common.h, line 28 at r1 (raw file):

Previously, janekmi (Jan Michalski) wrote…

Can we just make it a combination instead of using a hardcoded value?

Done.

---

tests/unit/peer/peer-mr_reg.c, line 40 at r1 (raw file):

Previously, janekmi (Jan Michalski) wrote…

Since the usage -> access calculation is now a part of the peer module I think also the testing of it should go here.
Moreover, we will need two protection domains: one iWARP and one non-iWARP to test both implementations.

Can it be done as a separate pull request?

[janekmi]

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

---

src/peer.c, line 33 at r2 (raw file):

/*
 * usage_to_access -- convert usage to access

Name.

---

src/peer.c, line 80 at r2 (raw file):

}

/* internal librpma API */

Bring it one function up. ;-)

---

tests/unit/peer/peer-mr_reg.c, line 40 at r1 (raw file):

Previously, ldorau (Lukasz Dorau) wrote…

Can it be done as a separate pull request?

I can agree to it if you will leave a big XXX somewhere. xD

[ldorau]

Reviewable status: 9 of 10 files reviewed, 3 unresolved discussions (waiting on @janekmi and @ldorau)

---

src/peer.c, line 33 at r2 (raw file):

Previously, janekmi (Jan Michalski) wrote…

Name.

Done.

---

src/peer.c, line 80 at r2 (raw file):

Previously, janekmi (Jan Michalski) wrote…

Bring it one function up. ;-)

Done.

[ldorau]

Reviewable status: 7 of 10 files reviewed, 3 unresolved discussions (waiting on @janekmi)

---

tests/unit/peer/peer-mr_reg.c, line 40 at r1 (raw file):

Previously, janekmi (Jan Michalski) wrote…

I can agree to it if you will leave a big XXX somewhere. xD

Done.

[janekmi]

Reviewed 3 of 3 files at r3.
Reviewable status: all files reviewed, 2 unresolved discussions (waiting on @ldorau)

---

tests/unit/peer/peer-mr_reg.c, line 166 at r3 (raw file):

	struct ibv_pd *mock_ibv_pd = MOCK_IBV_PD;

	int n_iters = sizeof(usage2access) / sizeof(struct usage2access_s);

These should come as prestates.

---

tests/unit/peer/peer-mr_reg.c, line 166 at r3 (raw file):

	struct ibv_pd *mock_ibv_pd = MOCK_IBV_PD;

	int n_iters = sizeof(usage2access) / sizeof(struct usage2access_s);

Having these, tests for rpma_mr_reg() are too extensive: https://github.com/pmem/rpma/blob/master/tests/unit/mr/mr-reg.c#L23

[ldorau]

Reviewable status: 8 of 11 files reviewed, 2 unresolved discussions (waiting on @janekmi and @ldorau)

---

tests/unit/peer/peer-mr_reg.c, line 166 at r3 (raw file):

Previously, janekmi (Jan Michalski) wrote…

These should come as prestates.

Done.

[ldorau]

Reviewable status: 8 of 12 files reviewed, 2 unresolved discussions (waiting on @janekmi)

---

tests/unit/peer/peer-mr_reg.c, line 166 at r3 (raw file):

Previously, janekmi (Jan Michalski) wrote…

Having these, tests for rpma_mr_reg() are too extensive: https://github.com/pmem/rpma/blob/master/tests/unit/mr/mr-reg.c#L23

Done.

[janekmi]

Reviewed 4 of 4 files at r4.
Reviewable status: all files reviewed, 4 unresolved discussions (waiting on @ldorau)

---

tests/unit/peer/peer-common.h, line 47 at r4 (raw file):

	int is_odp_capable;
	struct rpma_peer *peer;
};

Please put this structure just before setup__peer_prestates().

---

tests/unit/peer/peer-common.h, line 55 at r4 (raw file):

int teardown__peer(void **peer_ptr);

int setup__peer_prestates(void **in_out);

pprestate

---

tests/unit/peer/peer-common.h, line 56 at r4 (raw file):

int setup__peer_prestates(void **in_out);
int teardown__peer_prestates(void **peer_ptr);

.

---

tests/unit/peer/peer-common.c, line 81 at r4 (raw file):

 * Note:
 * - in - int is_odp_capable
 * - out - struct rpma_peer *

Copy-pasted inadequately.

[ldorau]

Reviewable status: 10 of 12 files reviewed, 4 unresolved discussions (waiting on @janekmi)

---

tests/unit/peer/peer-common.h, line 47 at r4 (raw file):

Previously, janekmi (Jan Michalski) wrote…

Please put this structure just before setup__peer_prestates().

Done.

---

tests/unit/peer/peer-common.h, line 55 at r4 (raw file):

Previously, janekmi (Jan Michalski) wrote…

pprestate

Done.

---

tests/unit/peer/peer-common.h, line 56 at r4 (raw file):

Previously, janekmi (Jan Michalski) wrote…

.

Done.

---

tests/unit/peer/peer-common.c, line 81 at r4 (raw file):

Previously, janekmi (Jan Michalski) wrote…

Copy-pasted inadequately.

It is not the same as in setup__peer() - I have changed it, but perhaps it will be better to remove it at all.

[janekmi]

Reviewed 2 of 2 files at r5.
Reviewable status: complete! all files reviewed, all discussions resolved (waiting on @ldorau)

[osalyk]

Reviewable status: complete! all files reviewed, all discussions resolved (waiting on @ldorau)