summaryrefslogtreecommitdiff
path: root/tools/perf/scripts/python
diff options
context:
space:
mode:
authorAshish Kumar Dhanotiya <adhanoti@codeaurora.org>2020-04-10 12:42:08 +0530
committerGerrit - the friendly Code Review server <code-review@localhost>2020-04-21 22:52:25 -0700
commit3f9b53f6d4f6ea4f712a45fa0b13d6fd9203335a (patch)
treed3a90549d9eca6c6c4aa1cc5f7b0fa47e781cb6f /tools/perf/scripts/python
parent11c68af53c75a11e23ede3b98d6f373b859244a9 (diff)
qcacld-3.0: Set power stats response cb and context to NULL
Currently for below cases there is a possiblity of mem leak in hdd_power_debugstats_cb: 1. If the power debug stats command is invoked with multiple threads from userspace, in this case the thread which invokes this command last, will over write the cookie in the sme context. Now when FW sends the responses for both the requests, hdd_power_debugstats_cb is invoked with the same cookie and in the call back processing there is memory allocated for the private data which will also be allocated twice and the second mem alloc will over write the pointer of private data from the memory allocated by the first mem alloc to second mem alloc. This will result in the memory leak for the first mem alloc. 2. If FW sends multiple responses for a single power debug stats command, there is a possiblity that memory can be allocated multiple times in hdd_power_debugstats_cb for debug registers as explained in point 1 and mem_free happens only once leading to memory leak for the other instances where memory is allocated. 3. Command gets invoked from userspace and cookie corresponding to this command is stored in sme. If this command gets timed-out user space thread is unblocked and if it again issues the same command, cookie for this command is stored in sme which will over write the cookie of the previous command. Now if FW sends the response of both the commands back to back, in this case the hdd_power_debugstats_cb will be invoked twice with the same cookie and will lead to mem leak as mentioned in point 1. To address above issue, hdd_power_debugstats_cb needs to be invoked only once for one cookie, set the callback and the cookie to NULL in sme after invoking the callback. Change-Id: Id473fd8f6530c57cc1096fd942a11aaf374c815d CRs-Fixed: 2664788
Diffstat (limited to 'tools/perf/scripts/python')
0 files changed, 0 insertions, 0 deletions