On 8/7/2024 10:59 PM, Suzuki K Poulose wrote:
On 18/07/2024 03:22, Tao Zhang wrote:On 7/11/2024 9:05 PM, Suzuki K Poulose wrote:On 11/07/2024 09:17, Tao Zhang wrote:In order to enable the output ports of multi-port output devices,This has nothing to do with multi-port. You wanted to pain the picture of a multi-port device, while the reality is different.such as static replicator, to correspond to designated sources, a mechanism for filtering data sources is introduced for the output ports. The specified source will be marked like below in the Devicetree. test-replicator { ... ... ... ... out-ports { ... ... ... ... port@0 { reg = <0>; xxx: endpoint { remote-endpoint = <&xxx>; filter_src = <&xxx>; <-- To specify the source to }; be filtered out here. }; port@1 { reg = <1>; yyy: endpoint { remote-endpoint = <&yyy>; filter_src = <&yyy>; <-- To specify the source to }; be filtered out here. }; }; };Then driver will find the expected source marked in the Devicetree, andsave it to the coresight path. When the function needs to filter the source, it could obtain it from coresight path parameter. Finally,the output port knows which source it corresponds to, and it also knowswhich input port it corresponds to.Minor nit: I think the commit description is full of "How" you are doing something, while it must rather be: "What and Why?"I would prefer something like : Subject: coresight: Add support for trace filtering by sourceSome replicators have hard coded filtering of "trace" data, based on thesource device. This is different from the trace filtering based on TraceID, available in the standard programmable replicators. e.g., Qualcomm replicators have filtering based on custom trace protocol format and is not programmable. The source device could be connected to the replicator via intermediate components (e.g., a funnel). Thus we need platform information from the firmware tables to decide the source device corresponding to a given output port from the replicator. Given this affects "trace path building" and traversing the path back from the sink to source, add the concept of "filtering by source" to the generic coresight connection.I will update according to your comments to the next version.Signed-off-by: Tao Zhang <quic_taozha@xxxxxxxxxxx> ---drivers/hwtracing/coresight/coresight-core.c | 125 ++++++++++++++++--.../hwtracing/coresight/coresight-platform.c | 18 +++ include/linux/coresight.h | 5 + 3 files changed, 135 insertions(+), 13 deletions(-)diff --git a/drivers/hwtracing/coresight/coresight-core.c b/drivers/hwtracing/coresight/coresight-core.cindex 9fc6f6b863e0..3f02a31b9328 100644 --- a/drivers/hwtracing/coresight/coresight-core.c +++ b/drivers/hwtracing/coresight/coresight-core.c@@ -75,15 +75,60 @@ struct coresight_device *coresight_get_percpu_sink(int cpu)} EXPORT_SYMBOL_GPL(coresight_get_percpu_sink);+static struct coresight_device *coresight_get_source(struct list_head *path)+{ + struct coresight_device *csdev; + + if (!path) + return NULL; ++ csdev = list_first_entry(path, struct coresight_node, link)->csdev;+ if (csdev->type != CORESIGHT_DEV_TYPE_SOURCE) + return NULL; + + return csdev; +} +Why is this still here ? Didn't agree to remove this and pass the sourcedirectly ?I agree pass the source directly, but how can we get the source device of the pathwithout this function? The solution I can think of is to pass source directly from"coresight_enable"like below.coresight_enable ---pass source---> coresight_enable_path -> coresight_enable_link-> coresight_find_out_connection -> coresight_blocks_sourceDoes this solution meet your expectations? Or do you have a better suggestion?+/**+ * coresight_source_filter - checks whether the connection matches the source+ * of path if connection is binded to specific source. + * @trace_source: The source device of the trace path + * @conn: The connection of one outport + *+ * Return zero if the connection doesn't have a source binded or source of the+ * path matches the source binds to connection. + */+static int coresight_source_filter(struct coresight_device *trace_source,+ struct coresight_connection *conn)This name is a bit confusing. It doesn't tell you, whether it allows the trace or blocks it.+{ + int ret = 0; + + if (!conn->filter_src_dev) + return ret;This is not the correct check. It must be : if (!conn->filter_src_fwnode)Because, the device could have disappeared (or not appeared yet). e.g., (TPDM) driver module unloaded/not - loaded.+ + if (!trace_source)Is that possible ?+ return -1; + + if (conn->filter_src_dev == trace_source) + ret = 0; + else + ret = -1;Couldn't this simply be : /* * Returns true, if the trace path is not possible through @conn * for trace originating from @src */ static bool coresight_blocks_source(src, conn) { return conn->filter_src_fwnode && (conn->filter_src_dev != src); }I will update to the next version.+ + return ret; +} + static struct coresight_connection * coresight_find_out_connection(struct coresight_device *src_dev, - struct coresight_device *dest_dev) + struct coresight_device *dest_dev, + struct coresight_device *trace_source)Please could you rename : src_dev => csdev dest_dev => out_dev trace_source => trace_src ? Having src_dev and trace_source in the same list is confusing.I will update to the next version.{ int i; struct coresight_connection *conn; for (i = 0; i < src_dev->pdata->nr_outconns; i++) { conn = src_dev->pdata->out_conns[i]; + if (coresight_source_filter(trace_source, conn))if (coresight_blocks_source(trace_source, conn)) continue;+ continue; if (conn->dest_dev == dest_dev) return conn; }@@ -251,7 +296,8 @@ static void coresight_disable_sink(struct coresight_device *csdev)static int coresight_enable_link(struct coresight_device *csdev, struct coresight_device *parent, - struct coresight_device *child) + struct coresight_device *child, + struct coresight_device *trace_source) { int link_subtype; struct coresight_connection *inconn, *outconn;@@ -259,8 +305,8 @@ static int coresight_enable_link(struct coresight_device *csdev,if (!parent || !child) return -EINVAL; - inconn = coresight_find_out_connection(parent, csdev); - outconn = coresight_find_out_connection(csdev, child);+ inconn = coresight_find_out_connection(parent, csdev, trace_source); + outconn = coresight_find_out_connection(csdev, child, trace_source);link_subtype = csdev->subtype.link_subtype;if (link_subtype == CORESIGHT_DEV_SUBTYPE_LINK_MERG && IS_ERR(inconn)) @@ -273,15 +319,16 @@ static int coresight_enable_link(struct coresight_device *csdev,static void coresight_disable_link(struct coresight_device *csdev, struct coresight_device *parent, - struct coresight_device *child) + struct coresight_device *child, + struct coresight_device *trace_source) { struct coresight_connection *inconn, *outconn; if (!parent || !child) return; - inconn = coresight_find_out_connection(parent, csdev); - outconn = coresight_find_out_connection(csdev, child);+ inconn = coresight_find_out_connection(parent, csdev, trace_source); + outconn = coresight_find_out_connection(csdev, child, trace_source);link_ops(csdev)->disable(csdev, inconn, outconn); }@@ -341,6 +388,9 @@ static void coresight_disable_path_from(struct list_head *path,{ u32 type; struct coresight_device *csdev, *parent, *child; + struct coresight_device *source; + + source = coresight_get_source(path);Grr! This must come from the "caller of the coresight_disable_path_from()", which was my comment. Please go backand double check the comment on previous version.How about if I pass the source to " coresight_disable_path_from" directly from"coresight_enable" or "coresight_disable" like below?coresight_enable ----| |---> coresight_disable_patch_from|---pass source---|coresight_disable----| |---> coresight_disable_path -> coresight_disable_patch_fromThis can ensure that the source passed into coresight_disable_patch_from isthe correct source.coresight_disable_path_from() must be given the "source" separately, as the "path" could be "partial". Rest could stay as it is in your patch.
In what scenario is this "path" "partial"?There are two place where "coresight_disable_path_from" will be called in coresight driver.
1. coresight_disable_path -> coresight_disable_path_fromIn this case, I see all "coresight_disable_path" will be called after building path successfully. The "path" should
be "complete", right?2. coresight_enable_path return with error and call coresight_disable_path_from to disable the "path" from
"nd". But in this case, it is called after building path completely, right? So this "path" should be complete as well?
Please let me know if there is any mistake in my understanding. If my understanding is correct, then there is no
need to pass "source" to "coresight_disable_path_from". Because the "source" can be collected from the "path"
directly.
if (!nd) nd = list_first_entry(path, struct coresight_node, link);@@ -375,7 +425,7 @@ static void coresight_disable_path_from(struct list_head *path,case CORESIGHT_DEV_TYPE_LINK: parent = list_prev_entry(nd, link)->csdev; child = list_next_entry(nd, link)->csdev; - coresight_disable_link(csdev, parent, child); + coresight_disable_link(csdev, parent, child, source); break; default: break;@@ -418,6 +468,9 @@ int coresight_enable_path(struct list_head *path, enum cs_mode mode,u32 type; struct coresight_node *nd; struct coresight_device *csdev, *parent, *child; + struct coresight_device *source; + + source = coresight_get_source(path); list_for_each_entry_reverse(nd, path, link) { csdev = nd->csdev;@@ -456,7 +509,7 @@ int coresight_enable_path(struct list_head *path, enum cs_mode mode,case CORESIGHT_DEV_TYPE_LINK: parent = list_prev_entry(nd, link)->csdev; child = list_next_entry(nd, link)->csdev; - ret = coresight_enable_link(csdev, parent, child);+ ret = coresight_enable_link(csdev, parent, child, source);if (ret) goto err; break;@@ -619,6 +672,7 @@ static void coresight_drop_device(struct coresight_device *csdev)* @csdev: The device to start from. * @sink: The final sink we want in this path. * @path: The list to add devices to. + * @trace_source: The trace source device of the path. * * The tree of Coresight device is traversed until @sink is found.* From there the sink is added to the list along with all the devices that led @@ -627,7 +681,8 @@ static void coresight_drop_device(struct coresight_device *csdev)*/ static int _coresight_build_path(struct coresight_device *csdev, struct coresight_device *sink, - struct list_head *path) + struct list_head *path, + struct coresight_device *trace_source) { int i, ret; bool found = false;@@ -639,7 +694,7 @@ static int _coresight_build_path(struct coresight_device *csdev, if (coresight_is_percpu_source(csdev) && coresight_is_percpu_sink(sink) && sink == per_cpu(csdev_sink, source_ops(csdev)->cpu_id(csdev))) {- if (_coresight_build_path(sink, sink, path) == 0) {+ if (_coresight_build_path(sink, sink, path, trace_source) == 0) {found = true; goto out; }@@ -650,8 +705,13 @@ static int _coresight_build_path(struct coresight_device *csdev,struct coresight_device *child_dev; child_dev = csdev->pdata->out_conns[i]->dest_dev; + + if (csdev->pdata->out_conns[i]->filter_src_dev+ && (csdev->pdata->out_conns[i]->filter_src_dev != trace_source))+ continue; + if (child_dev && - _coresight_build_path(child_dev, sink, path) == 0) {+ _coresight_build_path(child_dev, sink, path, trace_source) == 0) {found = true; break; }@@ -696,7 +756,7 @@ struct list_head *coresight_build_path(struct coresight_device *source,INIT_LIST_HEAD(path); - rc = _coresight_build_path(source, sink, path); + rc = _coresight_build_path(source, sink, path, source); if (rc) { kfree(path); return ERR_PTR(rc);@@ -957,6 +1017,14 @@ static int coresight_orphan_match(struct device *dev, void *data)/* This component still has an orphan */ still_orphan = true; }+ if ((conn->filter_src_fwnode) && dst_csdev + && (conn->filter_src_fwnode == dst_csdev->dev.fwnode)) { + conn->filter_src_dev = dst_csdev; + if (conn->filter_src_dev->type + != CORESIGHT_DEV_TYPE_SOURCE) + dev_warn(&conn->filter_src_dev->dev, + "Filter source is not a source device\n"); + }Have you tested your code with the following order : 1) Build everything as module 2) Load the replicator driver. 3) Load the TPDM driverAnd are you able to get the trace working ? I suspect it doesn't work. The csdev->orphan must be used to track if there is any missing linkwith unresolved "filter_src_dev".I really didn't consider this test case in my testing. There was a problem withthe logic of the code I modified. If it is modified to the following, can it solve thetest case you proposed? static int coresight_orphan_match(struct device *dev, void *data) { ... ... ... for (i = 0; i < src_csdev->pdata->nr_outconns; i++) { conn = src_csdev->pdata->out_conns[i];/*Add the following code to fix filter source device before skip the port*/+ if ((conn->filter_src_fwnode) && dst_csdev + && (conn->filter_src_fwnode == dst_csdev->dev.fwnode)) { + conn->filter_src_dev = dst_csdev; + if (conn->filter_src_dev->type + != CORESIGHT_DEV_TYPE_SOURCE) + dev_warn(&conn->filter_src_dev->dev, + "Filter source is not a source device\n");And we should not set the device as the filter_src_dev in this case.
Sure, I will update the code as the following.
+ if ((conn->filter_src_fwnode) && dst_csdev
+ && (conn->filter_src_fwnode == dst_csdev->dev.fwnode)) {
+ if (dst_csdev->type == CORESIGHT_DEV_TYPE_SOURCE)
+ conn->filter_src_dev = dst_csdev;
+ else
+ dev_warn(&dst_csdev->dev,
+ "Filter source is not a source device\n");
+ } /* Skip the port if it's already connected. */ if (conn->dest_dev) continue; ... ... ... ret = coresight_add_in_conn(conn); if (ret) return ret;/* Add the following code to make sure the device still orphan if there is a* filter source firmware node but the filter source is null.*/+ if (conn->filter_src_fwnode && !conn->filter_src_dev) + still_orphan = true; } else { /* This component still has an orphan */ still_orphan = true; } } src_csdev->orphan = still_orphan; ... ... ... }} src_csdev->orphan = still_orphan;@@ -974,6 +1042,30 @@ static int coresight_fixup_orphan_conns(struct coresight_device *csdev)csdev, coresight_orphan_match); }+/* reset_filter_src - Reset the filter source if the source is being removed */+static int reset_filter_src(struct coresight_device *trace_source, + struct coresight_connection *conn) +{ + int i; + int ret = -1; + struct coresight_device *csdev; + struct coresight_connection *output_conn;I don't think this is sufficient. I would rather walk through the entire coresight bus and fixup the filter_src_dev, rather than traversing thepaths. e.g., one of the devices in the path could have been removed. See coresight_clear_default_sink() for e.g. You should have something similar : coresight_clear_filter_source();That's right. My current understanding is that there are two cases that need to be considered.#1 Unregister the source deviceIn this case, we need to walk through the entire coresight bus and fixup the filter_src_dev#2 Unregister the replicatorIterate over all output connections to set filter_src_dev and filter_src_fwnode to NULL.Is this logically correct? I have attached my sample code below, please help review it.+static void coresight_clear_filter_source(struct device *dev, void *data)+{ + int i; + struct coresight_device *source = data; + struct coresight_device *csdev = to_coresight_device(dev); + + for (i = 0; i < csdev->pdata->nr_outconns; ++i) { + if (csdev->pdata->out_conns[i]->filter_src_dev == source) + csdev->pdata->out_conns[i]->filter_src_dev = NULL; + } +}That looks good to me./* coresight_remove_conns - Remove other device's references to this device */static void coresight_remove_conns(struct coresight_device *csdev) { int i, j; struct coresight_connection *conn;+ if (csdev->type == CORESIGHT_DEV_TYPE_SOURCE) //walk through the entire coresight bus and fixup the filter_src_dev+ bus_for_each_dev(&coresight_bustype, NULL, csdev, + coresight_clear_filter_source); /** Remove the input connection references from the destination device* for each output connection. */ for (i = 0; i < csdev->pdata->nr_outconns; i++) { conn = csdev->pdata->out_conns[i];+ if (conn->filter_src_fwnode) { //Iterate over all output connections to set filter_src_dev and filter_src_fwnode to NULL.+ conn->filter_src_dev = NULL; + conn->filter_src_fwnode = NULL;We may need to "drop" the refcount for this one ?
I'm not clear why need to "drop" the refcount here. I didn't add refcount to this patch series before.
Could you show me your sample code so I can understand what you mean? Best, Tao
Suzuki+ } if (!conn->dest_dev) continue; ... ... ... } Best, TaoSuzuki+ + csdev = conn->dest_dev; + for (i = 0; i < csdev->pdata->nr_outconns; i++) { + output_conn = csdev->pdata->out_conns[i]; + if (output_conn->filter_src_dev + && (output_conn->filter_src_dev == trace_source)) { + output_conn->filter_src_dev = NULL; + return 0; + } + ret = reset_filter_src(trace_source, output_conn); + if (!ret) + return ret; + } + return ret; +} +/* coresight_remove_conns - Remove other device's references to this device */static void coresight_remove_conns(struct coresight_device *csdev) {@@ -986,6 +1078,13 @@ static void coresight_remove_conns(struct coresight_device *csdev)*/for (i = 0; i < csdev->pdata->nr_outconns; i++) { conn = csdev->pdata->out_conns[i]; + if (csdev->type == CORESIGHT_DEV_TYPE_SOURCE) + reset_filter_src(csdev, conn); + if (conn->filter_src_fwnode) { + conn->filter_src_dev = NULL; + conn->filter_src_fwnode = NULL; + } + if (!conn->dest_dev) continue;diff --git a/drivers/hwtracing/coresight/coresight-platform.c b/drivers/hwtracing/coresight/coresight-platform.cindex 64e171eaad82..b3c3e2361d07 100644 --- a/drivers/hwtracing/coresight/coresight-platform.c +++ b/drivers/hwtracing/coresight/coresight-platform.c@@ -243,6 +243,24 @@ static int of_coresight_parse_endpoint(struct device *dev,conn.dest_fwnode = fwnode_handle_get(rdev_fwnode); conn.dest_port = rendpoint.port; + /* + * Get the firmware node of the filter source through the + * reference. This could be used to filter the source in + * building path. + */ + conn.filter_src_fwnode = + fwnode_find_reference(&ep->fwnode, "filter-src", 0); + if (IS_ERR(conn.filter_src_fwnode)) + conn.filter_src_fwnode = NULL; + else { + conn.filter_src_dev = + coresight_find_csdev_by_fwnode(conn.filter_src_fwnode); + if (conn.filter_src_dev && (conn.filter_src_dev->type + != CORESIGHT_DEV_TYPE_SOURCE)) + dev_warn(&conn.filter_src_dev->dev, + "Filter source %s is not a source device\n"); + } + new_conn = coresight_add_out_conn(dev, pdata, &conn); if (IS_ERR_VALUE(new_conn)) { fwnode_handle_put(conn.dest_fwnode); diff --git a/include/linux/coresight.h b/include/linux/coresight.h index f09ace92176e..91a689b4514b 100644 --- a/include/linux/coresight.h +++ b/include/linux/coresight.h @@ -172,6 +172,9 @@ struct coresight_desc { * @dest_dev: a @coresight_device representation of the component connected to @src_port. NULL until the device is created * @link: Representation of the connection as a sysfs link. + * @filter_src_fwnode: filter source component's fwnode handle.+ * @filter_src_dev: a @coresight_device representation of the component that+ needs to be filtered. ** The full connection structure looks like this, where in_conns store* references to same connection as the source device's out_conns. @@ -200,6 +203,8 @@ struct coresight_connection { struct coresight_device *dest_dev; struct coresight_sysfs_link *link; struct coresight_device *src_dev; + struct fwnode_handle *filter_src_fwnode; + struct coresight_device *filter_src_dev; atomic_t src_refcnt; atomic_t dest_refcnt; };
