** Description changed:
[Impact]
The functional change is negligible - the primary goal here is to align our
tree closer to upstream to make future backports cleaner (and therefore safer).
- We are currently carrying 61 SAUCE patches for the hns3 driver. All but
- 6 of these have gone upstream in some form. In one case, the upstream
- patch corrected a mistake in the SAUCE counterpart:
+ We are currently carrying 61 SAUCE patches for the hns3 driver. 5 of
+ these changes haven't been posted yet, 1 has been rejected, but the
+ remaining have now all been merged.
+
+ Functionally, the upstream patches differ from our SAUCE in 2 ways:
+
+ 1) In one case, the upstream patch corrected a mistake in the SAUCE
+ counterpart:
e22b531b5889d UBUNTU: SAUCE: {topost} net: hns3: modify hnae_ to hnae3_
While it purports to just be a symbol rename, it also made the following
change (the test_bit() addition), which I've verified with upstream is
incorrect:
@@ -2703,7 +2703,8 @@ static int hclge_reset_wait(struct hclge_dev *hdev)
}
val = hclge_read_dev(&hdev->hw, reg);
- while (hnae_get_bit(val, reg_bit) && cnt < HCLGE_RESET_WAIT_CNT) {
+ while (hnae3_get_bit(val, reg_bit) && cnt < HCLGE_RESET_WAIT_CNT &&
+ test_bit(HCLGE_STATE_RST_HANDLING, &hdev->state)) {
msleep(HCLGE_RESET_WATI_MS);
val = hclge_read_dev(&hdev->hw, reg);
cnt++;
+ 2) One patch was rejected for its performance impact, and Ubuntu should
+ follow suit:
+
+ c4a05b6f9f4b6 UBUNTU: SAUCE: {topost} net: hns3: using modulo for cyclic
+ counters in hclge_cmd_send
+
+ See: https://lore.kernel.org/patchwork/patch/957111/
+
In addition, several of the patches have changed in non-functional ways
(whitespace, comments, commit messages, etc). This will likely prevent
future patches from cherry-picking cleanly and, in the case of commit
messages, obfuscate which fixes we already have.
[Test Case]
- 1) git diff, to make sure delta after includes only the test_bit() correction
and non-functional changes.
+ 1) git diff, to make sure delta after includes only the test_bit()
correction, modulo removal, and non-functional changes.
2) Functional test on D06 hardware (the only impacted platform).
[Fix]
- Revert all the SAUCE patches that have countparts upstream, and apply the
upstream cherry-picks instead.
+ Revert all the SAUCE patches that have counterparts upstream, and apply the
upstream cherry-picks instead. Revert the rejected patch.
[Regression Risk]
- The git diff from before and after this series clearly shows only the one
correction and the various cosmetic changes.
+ The git diff from before and after this series clearly shows only the one
correction, the one revert, and the various cosmetic changes.
--
You received this bug notification because you are a member of Kernel
Packages, which is subscribed to linux in Ubuntu.
https://bugs.launchpad.net/bugs/1787477
Title:
Reconcile hns3 SAUCE patches with upstream
Status in linux package in Ubuntu:
In Progress
Status in linux source package in Bionic:
In Progress
Bug description:
[Impact]
The functional change is negligible - the primary goal here is to align our
tree closer to upstream to make future backports cleaner (and therefore safer).
We are currently carrying 61 SAUCE patches for the hns3 driver. 5 of
these changes haven't been posted yet, 1 has been rejected, but the
remaining have now all been merged.
Functionally, the upstream patches differ from our SAUCE in 2 ways:
1) In one case, the upstream patch corrected a mistake in the SAUCE
counterpart:
e22b531b5889d UBUNTU: SAUCE: {topost} net: hns3: modify hnae_ to
hnae3_
While it purports to just be a symbol rename, it also made the
following change (the test_bit() addition), which I've verified with
upstream is incorrect:
@@ -2703,7 +2703,8 @@ static int hclge_reset_wait(struct hclge_dev *hdev)
}
val = hclge_read_dev(&hdev->hw, reg);
- while (hnae_get_bit(val, reg_bit) && cnt < HCLGE_RESET_WAIT_CNT) {
+ while (hnae3_get_bit(val, reg_bit) && cnt < HCLGE_RESET_WAIT_CNT &&
+ test_bit(HCLGE_STATE_RST_HANDLING, &hdev->state)) {
msleep(HCLGE_RESET_WATI_MS);
val = hclge_read_dev(&hdev->hw, reg);
cnt++;
2) One patch was rejected for its performance impact, and Ubuntu
should follow suit:
c4a05b6f9f4b6 UBUNTU: SAUCE: {topost} net: hns3: using modulo for
cyclic counters in hclge_cmd_send
See: https://lore.kernel.org/patchwork/patch/957111/
In addition, several of the patches have changed in non-functional
ways (whitespace, comments, commit messages, etc). This will likely
prevent future patches from cherry-picking cleanly and, in the case of
commit messages, obfuscate which fixes we already have.
[Test Case]
1) git diff, to make sure delta after includes only the test_bit()
correction, modulo removal, and non-functional changes.
2) Functional test on D06 hardware (the only impacted platform).
[Fix]
Revert all the SAUCE patches that have counterparts upstream, and apply the
upstream cherry-picks instead. Revert the rejected patch.
[Regression Risk]
The git diff from before and after this series clearly shows only the one
correction, the one revert, and the various cosmetic changes.
To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1787477/+subscriptions
--
Mailing list: https://launchpad.net/~kernel-packages
Post to : kernel-packages@lists.launchpad.net
Unsubscribe : https://launchpad.net/~kernel-packages
More help : https://help.launchpad.net/ListHelp
