(mynewt-core) branch master updated: hw/bus: Update sanity check for locking

[jerzy]

This is an automated email from the ASF dual-hosted git repository.

jerzy pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/mynewt-core.git


The following commit(s) were added to refs/heads/master by this push:
     new 793e95ceb hw/bus: Update sanity check for locking
793e95ceb is described below

commit 793e95ceb7f790263c7f2ab2a6818315aaf45b19
Author: Jerzy Kasenberg <jerzy.kasenb...@codecoup.pl>
AuthorDate: Sat Oct 11 21:09:03 2025 +0200

    hw/bus: Update sanity check for locking
    
    Code assume that node can be locked only once and
    second lock is probably for another node.
    Lock are internally use on each transaction.
    Write, read, write then read transactions can be
    used with just one lock.
    spiflash driver locks node when it wants to write,
    however writing consist of two transactions fist
    command is sent (4 bytes) then another write is
    done with separate buffer. CS must be low during
    those two writes. While write could be performed
    successfully current implementation would require
    SPI controller to be unlocked between writes.
    This could potentially lead to SPI controller
    being locked for another device while CS is still
    down for spi flash.
    In fact due to sanity check spi flash driver
    did not work at all with any bus driver since
    code tried to lock device before writes and
    then both writes would fail due to nested lock
    sanity check.
    
    Now sanit check verifies that nested lock is
    for same device.
    
    Signed-off-by: Jerzy Kasenberg <je...@apache.org>
---
 hw/bus/src/bus.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/hw/bus/src/bus.c b/hw/bus/src/bus.c
index e764bfcb9..cbf368bb7 100644
--- a/hw/bus/src/bus.c
+++ b/hw/bus/src/bus.c
@@ -531,11 +531,13 @@ bus_node_lock(struct os_dev *node, os_time_t timeout)
 #endif
 
     /*
-     * Configuration is done on 1st lock so in case we need to configure device
-     * on nested lock it means that most likely bus device was locked for one
-     * node and then access is done on another node which is not correct.
+     * Configuration is done on 1st lock. For nested lock make sure that
+     * bnode did not change.
      */
     if (MYNEWT_VAL(OS_SCHEDULING) && os_mutex_get_level(&bdev->lock) != 1) {
+        if (bdev->configured_for == bnode) {
+            return 0;
+        }
         (void)bus_node_unlock(node);
         return SYS_EACCES;
     }