| Message ID | 20220303223351.141238-1-bjorn.andersson@linaro.org (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
| Series | [v3,1/6] device property: Helper to match multiple connections | expand |
On Thu, Mar 03, 2022 at 02:33:46PM -0800, Bjorn Andersson wrote: > In some cases multiple connections with the same connection id > needs to be resolved from a fwnode graph. > > One such example is when separate hardware is used for performing muxing > and/or orientation switching of the SuperSpeed and SBU lines in a USB > Type-C connector. In this case the connector needs to belong to a graph > with multiple matching remote endpoints, and the Type-C controller needs > to be able to resolve them both. > > Add a new API that allows this kind of lookup. ... > +static unsigned int fwnode_graph_devcon_matches(struct fwnode_handle *fwnode, > + const char *con_id, void *data, > + devcon_match_fn_t match, > + void **matches, > + unsigned int matches_len) > +{ > + struct fwnode_handle *node; > + struct fwnode_handle *ep; > + unsigned int count = 0; > + void *ret; > + > + fwnode_graph_for_each_endpoint(fwnode, ep) { > + if (count >= matches_len && matches) { > + fwnode_handle_put(ep); > + return count; > + } Wouldn't be the same as if (count >= matches_len) { fwnode_handle_put(ep); break; } ? > + node = fwnode_graph_get_remote_port_parent(ep); > + if (!fwnode_device_is_available(node)) { > + fwnode_handle_put(node); > + continue; > + } > + > + ret = match(node, con_id, data); > + fwnode_handle_put(node); > + if (ret) { > + if (matches) > + matches[count] = ret; > + count++; > + } > + } > + return count; > +} ... > +static unsigned int fwnode_devcon_matches(struct fwnode_handle *fwnode, > + const char *con_id, void *data, > + devcon_match_fn_t match, > + void **matches, > + unsigned int matches_len) > +{ > + struct fwnode_handle *node; > + unsigned int count = 0; > + unsigned int i; > + void *ret; > + > + for (i = 0; ; i++) { > + if (count >= matches_len && matches) > + return count; Ditto. > + node = fwnode_find_reference(fwnode, con_id, i); > + if (IS_ERR(node)) > + break; > + > + ret = match(node, NULL, data); > + fwnode_handle_put(node); > + if (ret) { > + if (matches) > + matches[count] = ret; > + count++; > + } > + } > + > + return count; > +} ... > +int fwnode_connection_find_matches(struct fwnode_handle *fwnode, > + const char *con_id, void *data, > + devcon_match_fn_t match, > + void **matches, unsigned int matches_len) > +{ > + unsigned int count_graph; > + unsigned int count_ref; > + > + if (!fwnode || !match) > + return -EINVAL; > + count_graph = fwnode_graph_devcon_matches(fwnode, con_id, data, match, > + matches, matches_len); > + matches += count_graph; > + matches_len -= count_graph; No, won't work when matches == NULL. Also, matches_len is expected to be 0 in that case (or at least being ignored, check with vsnprintf() behaviour in similar case). So, something like this, perhaps if (matches && matches_len) { matches += count_graph; matches_len -= count_graph; } > + count_ref = fwnode_devcon_matches(fwnode, con_id, data, match, > + matches, matches_len); > + > + return count_graph + count_ref; > +}
Hi Andy, Björn, On Fri, Mar 04, 2022 at 02:54:21PM +0200, Andy Shevchenko wrote: > On Thu, Mar 03, 2022 at 02:33:46PM -0800, Bjorn Andersson wrote: > > In some cases multiple connections with the same connection id > > needs to be resolved from a fwnode graph. > > > > One such example is when separate hardware is used for performing muxing > > and/or orientation switching of the SuperSpeed and SBU lines in a USB > > Type-C connector. In this case the connector needs to belong to a graph > > with multiple matching remote endpoints, and the Type-C controller needs > > to be able to resolve them both. > > > > Add a new API that allows this kind of lookup. > > ... > > > +static unsigned int fwnode_graph_devcon_matches(struct fwnode_handle *fwnode, > > + const char *con_id, void *data, > > + devcon_match_fn_t match, > > + void **matches, > > + unsigned int matches_len) > > +{ > > + struct fwnode_handle *node; > > + struct fwnode_handle *ep; > > + unsigned int count = 0; > > + void *ret; > > + > > + fwnode_graph_for_each_endpoint(fwnode, ep) { > > > + if (count >= matches_len && matches) { > > + fwnode_handle_put(ep); > > + return count; > > + } > > Wouldn't be the same as > > if (count >= matches_len) { > fwnode_handle_put(ep); > break; > } Don't you need to check for non-NULL matches here? I find return above easier to read. > > ? > > > + node = fwnode_graph_get_remote_port_parent(ep); > > + if (!fwnode_device_is_available(node)) { > > + fwnode_handle_put(node); > > + continue; > > + } > > + > > + ret = match(node, con_id, data); > > + fwnode_handle_put(node); > > + if (ret) { > > + if (matches) > > + matches[count] = ret; > > + count++; > > + } > > + } > > + return count; > > +} > > ... > > > +static unsigned int fwnode_devcon_matches(struct fwnode_handle *fwnode, > > + const char *con_id, void *data, > > + devcon_match_fn_t match, > > + void **matches, > > + unsigned int matches_len) > > +{ > > + struct fwnode_handle *node; > > + unsigned int count = 0; > > + unsigned int i; > > + void *ret; > > + > > + for (i = 0; ; i++) { > > > + if (count >= matches_len && matches) > > + return count; > > Ditto. > > > + node = fwnode_find_reference(fwnode, con_id, i); > > + if (IS_ERR(node)) > > + break; > > + > > + ret = match(node, NULL, data); > > + fwnode_handle_put(node); > > + if (ret) { > > + if (matches) > > + matches[count] = ret; > > + count++; > > + } > > + } > > + > > + return count; > > +} > > ... > > > +int fwnode_connection_find_matches(struct fwnode_handle *fwnode, > > + const char *con_id, void *data, > > + devcon_match_fn_t match, > > + void **matches, unsigned int matches_len) > > +{ > > + unsigned int count_graph; > > + unsigned int count_ref; > > + > > + if (!fwnode || !match) > > + return -EINVAL; > > > + count_graph = fwnode_graph_devcon_matches(fwnode, con_id, data, match, > > + matches, matches_len); > > > + matches += count_graph; > > + matches_len -= count_graph; > > No, won't work when matches == NULL. > > Also, matches_len is expected to be 0 in that case (or at least being ignored, > check with vsnprintf() behaviour in similar case). > > So, something like this, perhaps > > if (matches && matches_len) { > matches += count_graph; > matches_len -= count_graph; > } Good find! > > > + count_ref = fwnode_devcon_matches(fwnode, con_id, data, match, > > + matches, matches_len); > > + > > + return count_graph + count_ref; > > +} > >
On Fri, Mar 04, 2022 at 03:41:33PM +0200, Sakari Ailus wrote: > On Fri, Mar 04, 2022 at 02:54:21PM +0200, Andy Shevchenko wrote: > > On Thu, Mar 03, 2022 at 02:33:46PM -0800, Bjorn Andersson wrote: ... > > > + if (count >= matches_len && matches) { > > > + fwnode_handle_put(ep); > > > + return count; > > > + } > > > > Wouldn't be the same as > > > > if (count >= matches_len) { > > fwnode_handle_put(ep); > > break; > > } > > Don't you need to check for non-NULL matches here? Right, this should be kept as in original patch. > I find return above easier to read. Okay, original code may work, so I have no strong opinion about return vs break, although I find slightly better to have a single point of return in such case. > > ? ... > > > + count_graph = fwnode_graph_devcon_matches(fwnode, con_id, data, match, > > > + matches, matches_len); > > > > > + matches += count_graph; > > > + matches_len -= count_graph; > > > > No, won't work when matches == NULL. > > > > Also, matches_len is expected to be 0 in that case (or at least being ignored, > > check with vsnprintf() behaviour in similar case). > > > > So, something like this, perhaps > > > > if (matches && matches_len) { > > matches += count_graph; > > matches_len -= count_graph; > > } > > Good find! I have checked vsnprintf() and indeed, it expects to have the size is 0 when the resulting buffer pointer is NULL, and it doesn't do any additional checks. > > > + count_ref = fwnode_devcon_matches(fwnode, con_id, data, match, > > > + matches, matches_len);
On Fri 04 Mar 06:54 CST 2022, Andy Shevchenko wrote: > On Thu, Mar 03, 2022 at 02:33:46PM -0800, Bjorn Andersson wrote: > > In some cases multiple connections with the same connection id > > needs to be resolved from a fwnode graph. > > > > One such example is when separate hardware is used for performing muxing > > and/or orientation switching of the SuperSpeed and SBU lines in a USB > > Type-C connector. In this case the connector needs to belong to a graph > > with multiple matching remote endpoints, and the Type-C controller needs > > to be able to resolve them both. > > > > Add a new API that allows this kind of lookup. > > ... > > > +static unsigned int fwnode_graph_devcon_matches(struct fwnode_handle *fwnode, > > + const char *con_id, void *data, > > + devcon_match_fn_t match, > > + void **matches, > > + unsigned int matches_len) > > +{ > > + struct fwnode_handle *node; > > + struct fwnode_handle *ep; > > + unsigned int count = 0; > > + void *ret; > > + > > + fwnode_graph_for_each_endpoint(fwnode, ep) { > > > + if (count >= matches_len && matches) { > > + fwnode_handle_put(ep); > > + return count; > > + } > > Wouldn't be the same as > > if (count >= matches_len) { This would cause the return value to be at most matches_len, seems relevant to ignore (or perhaps require it to be 0) when matches is NULL. But flipping the order of the expression seems better to me, now that this has been sitting for a while. > fwnode_handle_put(ep); > break; Right, this isn't an "early return", so nicer to have a single return at the bottom. > } > > ? > > > + node = fwnode_graph_get_remote_port_parent(ep); > > + if (!fwnode_device_is_available(node)) { > > + fwnode_handle_put(node); > > + continue; > > + } > > + > > + ret = match(node, con_id, data); > > + fwnode_handle_put(node); > > + if (ret) { > > + if (matches) > > + matches[count] = ret; > > + count++; > > + } > > + } > > + return count; > > +} > > ... > > > +static unsigned int fwnode_devcon_matches(struct fwnode_handle *fwnode, > > + const char *con_id, void *data, > > + devcon_match_fn_t match, > > + void **matches, > > + unsigned int matches_len) > > +{ > > + struct fwnode_handle *node; > > + unsigned int count = 0; > > + unsigned int i; > > + void *ret; > > + > > + for (i = 0; ; i++) { > > > + if (count >= matches_len && matches) > > + return count; > > Ditto. > > > + node = fwnode_find_reference(fwnode, con_id, i); > > + if (IS_ERR(node)) > > + break; > > + > > + ret = match(node, NULL, data); > > + fwnode_handle_put(node); > > + if (ret) { > > + if (matches) > > + matches[count] = ret; > > + count++; > > + } > > + } > > + > > + return count; > > +} > > ... > > > +int fwnode_connection_find_matches(struct fwnode_handle *fwnode, > > + const char *con_id, void *data, > > + devcon_match_fn_t match, > > + void **matches, unsigned int matches_len) > > +{ > > + unsigned int count_graph; > > + unsigned int count_ref; > > + > > + if (!fwnode || !match) > > + return -EINVAL; > > > + count_graph = fwnode_graph_devcon_matches(fwnode, con_id, data, match, > > + matches, matches_len); > > > + matches += count_graph; > > + matches_len -= count_graph; > > No, won't work when matches == NULL. > Sorry about that, you're obviously correct. > Also, matches_len is expected to be 0 in that case (or at least being ignored, > check with vsnprintf() behaviour in similar case). > > So, something like this, perhaps > > if (matches && matches_len) { > matches += count_graph; > matches_len -= count_graph; > } Seems reasonable. Thanks, Bjorn > > > + count_ref = fwnode_devcon_matches(fwnode, con_id, data, match, > > + matches, matches_len); > > + > > + return count_graph + count_ref; > > +} > > > -- > With Best Regards, > Andy Shevchenko > >
On Fri 04 Mar 07:52 CST 2022, Andy Shevchenko wrote: > On Fri, Mar 04, 2022 at 03:41:33PM +0200, Sakari Ailus wrote: > > On Fri, Mar 04, 2022 at 02:54:21PM +0200, Andy Shevchenko wrote: > > > On Thu, Mar 03, 2022 at 02:33:46PM -0800, Bjorn Andersson wrote: > > ... > > > > > + if (count >= matches_len && matches) { > > > > + fwnode_handle_put(ep); > > > > + return count; > > > > + } > > > > > > Wouldn't be the same as > > > > > > if (count >= matches_len) { > > > fwnode_handle_put(ep); > > > break; > > > } > > > > Don't you need to check for non-NULL matches here? > > Right, this should be kept as in original patch. > > > I find return above easier to read. > > Okay, original code may work, so I have no strong opinion about return vs > break, although I find slightly better to have a single point of return in > such case. > I like using early returns when possible, but this is not an early return and it is in the loop so it makes more sense to me to break out. > > > ? > > ... > > > > > + count_graph = fwnode_graph_devcon_matches(fwnode, con_id, data, match, > > > > + matches, matches_len); > > > > > > > + matches += count_graph; > > > > + matches_len -= count_graph; > > > > > > No, won't work when matches == NULL. > > > > > > Also, matches_len is expected to be 0 in that case (or at least being ignored, > > > check with vsnprintf() behaviour in similar case). > > > > > > So, something like this, perhaps > > > > > > if (matches && matches_len) { > > > matches += count_graph; > > > matches_len -= count_graph; > > > } > > > > Good find! > > I have checked vsnprintf() and indeed, it expects to have the size is 0 when > the resulting buffer pointer is NULL, and it doesn't do any additional checks. > Per the vsnprintf() semantics it's not the destination buffer being NULL that's significant, but rather just the length being 0 that matters. To follow that, I should fill @matches_len entries in @matches and then just continue counting without storing anything in @matches. But that won't work in this case, because in the event that the @match function returns something that has to be freed (such as the refcounted objects returned by the typec_mux code), dropping this in favor of just counting it would cause memory/reference leaks. As such, I think this should differ in that @matches = NULL is significant, and it's nice to not have matches_len turn negative/bogus in the count case. So I like your suggestion. Thanks, Bjorn > > > > + count_ref = fwnode_devcon_matches(fwnode, con_id, data, match, > > > > + matches, matches_len); > > -- > With Best Regards, > Andy Shevchenko > >
On Sun, Mar 06, 2022 at 08:18:24PM -0600, Bjorn Andersson wrote: > On Fri 04 Mar 07:52 CST 2022, Andy Shevchenko wrote: > > On Fri, Mar 04, 2022 at 03:41:33PM +0200, Sakari Ailus wrote: ... > > I have checked vsnprintf() and indeed, it expects to have the size is 0 when > > the resulting buffer pointer is NULL, and it doesn't do any additional checks. > > Per the vsnprintf() semantics it's not the destination buffer being NULL > that's significant, but rather just the length being 0 that matters. Actually the main point is to have buffer to be NULL. But at the same time size also has to be 0. When the size is 0, but buffer is not NULL, it's a different story. > To follow that, I should fill @matches_len entries in @matches and then > just continue counting without storing anything in @matches. > > But that won't work in this case, because in the event that the @match > function returns something that has to be freed (such as the refcounted > objects returned by the typec_mux code), dropping this in favor of just > counting it would cause memory/reference leaks. > > As such, I think this should differ in that @matches = NULL is > significant, and it's nice to not have matches_len turn negative/bogus > in the count case. > > So I like your suggestion.
diff --git a/drivers/base/property.c b/drivers/base/property.c index c0e94cce9c29..5cda205136f6 100644 --- a/drivers/base/property.c +++ b/drivers/base/property.c @@ -1218,6 +1218,40 @@ fwnode_graph_devcon_match(struct fwnode_handle *fwnode, const char *con_id, return NULL; } +static unsigned int fwnode_graph_devcon_matches(struct fwnode_handle *fwnode, + const char *con_id, void *data, + devcon_match_fn_t match, + void **matches, + unsigned int matches_len) +{ + struct fwnode_handle *node; + struct fwnode_handle *ep; + unsigned int count = 0; + void *ret; + + fwnode_graph_for_each_endpoint(fwnode, ep) { + if (count >= matches_len && matches) { + fwnode_handle_put(ep); + return count; + } + + node = fwnode_graph_get_remote_port_parent(ep); + if (!fwnode_device_is_available(node)) { + fwnode_handle_put(node); + continue; + } + + ret = match(node, con_id, data); + fwnode_handle_put(node); + if (ret) { + if (matches) + matches[count] = ret; + count++; + } + } + return count; +} + static void * fwnode_devcon_match(struct fwnode_handle *fwnode, const char *con_id, void *data, devcon_match_fn_t match) @@ -1240,6 +1274,37 @@ fwnode_devcon_match(struct fwnode_handle *fwnode, const char *con_id, return NULL; } +static unsigned int fwnode_devcon_matches(struct fwnode_handle *fwnode, + const char *con_id, void *data, + devcon_match_fn_t match, + void **matches, + unsigned int matches_len) +{ + struct fwnode_handle *node; + unsigned int count = 0; + unsigned int i; + void *ret; + + for (i = 0; ; i++) { + if (count >= matches_len && matches) + return count; + + node = fwnode_find_reference(fwnode, con_id, i); + if (IS_ERR(node)) + break; + + ret = match(node, NULL, data); + fwnode_handle_put(node); + if (ret) { + if (matches) + matches[count] = ret; + count++; + } + } + + return count; +} + /** * fwnode_connection_find_match - Find connection from a device node * @fwnode: Device node with the connection @@ -1267,3 +1332,45 @@ void *fwnode_connection_find_match(struct fwnode_handle *fwnode, return fwnode_devcon_match(fwnode, con_id, data, match); } EXPORT_SYMBOL_GPL(fwnode_connection_find_match); + +/** + * fwnode_connection_find_matches - Find connections from a device node + * @fwnode: Device node with the connection + * @con_id: Identifier for the connection + * @data: Data for the match function + * @match: Function to check and convert the connection description + * @matches: Array of pointers to fill with matches + * @matches_len: Length of @matches + * + * Find up to @matches_len connections with unique identifier @con_id between + * @fwnode and other device nodes. @match will be used to convert the + * connection description to data the caller is expecting to be returned + * through the @matches array. + * If @matches is NULL @matches_len is ignored and the total number of resolved + * matches is returned. + * + * Return: Number of matches resolved, or negative errno. + */ +int fwnode_connection_find_matches(struct fwnode_handle *fwnode, + const char *con_id, void *data, + devcon_match_fn_t match, + void **matches, unsigned int matches_len) +{ + unsigned int count_graph; + unsigned int count_ref; + + if (!fwnode || !match) + return -EINVAL; + + count_graph = fwnode_graph_devcon_matches(fwnode, con_id, data, match, + matches, matches_len); + + matches += count_graph; + matches_len -= count_graph; + + count_ref = fwnode_devcon_matches(fwnode, con_id, data, match, + matches, matches_len); + + return count_graph + count_ref; +} +EXPORT_SYMBOL_GPL(fwnode_connection_find_matches); diff --git a/include/linux/property.h b/include/linux/property.h index 4cd4b326941f..de7ff336d2c8 100644 --- a/include/linux/property.h +++ b/include/linux/property.h @@ -447,6 +447,11 @@ static inline void *device_connection_find_match(struct device *dev, return fwnode_connection_find_match(dev_fwnode(dev), con_id, data, match); } +int fwnode_connection_find_matches(struct fwnode_handle *fwnode, + const char *con_id, void *data, + devcon_match_fn_t match, + void **matches, unsigned int matches_len); + /* -------------------------------------------------------------------------- */ /* Software fwnode support - when HW description is incomplete or missing */
In some cases multiple connections with the same connection id needs to be resolved from a fwnode graph. One such example is when separate hardware is used for performing muxing and/or orientation switching of the SuperSpeed and SBU lines in a USB Type-C connector. In this case the connector needs to belong to a graph with multiple matching remote endpoints, and the Type-C controller needs to be able to resolve them both. Add a new API that allows this kind of lookup. Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org> --- Changes since v2: - Allow the caller of the new api to pass a matches of NULL, to count possible matches. I previously argued that this will cause memory leaks, but Andy pointed out that this depends on the caller and the match function. - Fixed spelling mistakes in commit message and kernel-doc. - Use two "count" variables to make the math clearer. Changes since v1: - Iterator in fwnode_devcon_matches() is now unsigned. - fwnode_handle_put() node for unavailable nodes. - Extended commit message on the subject of supporting dynamically sized "matches" array. drivers/base/property.c | 107 +++++++++++++++++++++++++++++++++++++++ include/linux/property.h | 5 ++ 2 files changed, 112 insertions(+)