summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorKay Sievers <kay@vrfy.org>2012-08-03 10:52:24 +0200
committerKay Sievers <kay@vrfy.org>2012-08-03 10:52:24 +0200
commit1c7dfbf249f5320d0675fcb03556a55881e21457 (patch)
tree33a7c6322ffffafb033738e318b7fa38e1d82429
parent3bea51c59d79333a2a365f4677c97b4302377a48 (diff)
udev: path_id - add comment about the unrteliable rebase logic
-rw-r--r--src/udev/udev-builtin-path_id.c18
1 files changed, 17 insertions, 1 deletions
diff --git a/src/udev/udev-builtin-path_id.c b/src/udev/udev-builtin-path_id.c
index 8eac991d3c..720f616da5 100644
--- a/src/udev/udev-builtin-path_id.c
+++ b/src/udev/udev-builtin-path_id.c
@@ -236,7 +236,23 @@ static struct udev_device *handle_scsi_default(struct udev_device *parent, char
if (sscanf(name, "%d:%d:%d:%d", &host, &bus, &target, &lun) != 4)
return NULL;
- /* rebase host offset to get the local relative number */
+ /*
+ * Rebase host offset to get the local relative number
+ *
+ * Note: This is by definition racy, unreliable and too simple.
+ * Please do not copy this model anywhere. It's just a left-over
+ * from the time we had no idea how things should look like in
+ * the end.
+ *
+ * Making assumptions about a global in-kernel counter and use
+ * that to calculate a local offset is a very broken concept. It
+ * can only work as long as things are in strict order.
+ *
+ * The kernel needs to export the instance/port number of a
+ * controller directly, without the need for rebase magic like
+ * this. Manual driver unbind/bind, parallel hotplug/unplug will
+ * get into the way of this "I hope it works" logic.
+ */
basenum = -1;
base = strdup(udev_device_get_syspath(hostdev));
if (base == NULL)