eventdev: prefix global variable
Checks
Commit Message
A global defined with name "stats".
DPDK is a library, please don't define global variables with this
generic name.
Fixes: b1ce8ebd97ba ("eventdev: add PMD callbacks for eth Rx adapter")
Cc: stable@dpdk.org
Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
---
There are already questions about that "stats" variable:
- Why a global variable defined in header file?
- It seems it is not used at all, remove it?
Forgotten to be removed in below commit, but because of generic
"stats" name, it didn't cause any compile error but happily used
global variable
Fixes: 3810ae435783 ("eventdev: add interrupt driven queues to Rx adapter")
But this patch mainly sent as an RFC, to ask what would you think about
a cleanup patch to add the library prefix for all the global variables
that are defined in libries? (yes we have lots of them)
Like:
zip_cc --> octeontx_zip_cc [1]
valid_args --> qede_valid_args [2]
scheduler --> crypto_scheduler [3]
[1]
Argh! This also looks wrong, I think intention was make this enum
name instead of global variable, can you please check. (This was random
sample I chosed.)
Commit 43e610bb8565 ("compress/octeontx: introduce octeontx zip PMD")
Cc: sunila.sahu@caviumnetworks.com
[2]
This is also wrong, it should be static,
Commit f64b91b0eb5d ("net/qede: replace config option with run-time arg")
Cc: rasesh.mody@cavium.com
[3]
This should be static too,
Commit 100e4f7e44ab ("crypto/scheduler: add round-robin mode")
Cc: roy.fan.zhang@intel.com
---
lib/librte_eventdev/rte_event_eth_rx_adapter.c | 2 +-
lib/librte_eventdev/rte_eventdev_pmd.h | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
Comments
On 10/3/2018 2:48 PM, Ferruh Yigit wrote:
> A global defined with name "stats".
>
> DPDK is a library, please don't define global variables with this
> generic name.
>
> Fixes: b1ce8ebd97ba ("eventdev: add PMD callbacks for eth Rx adapter")
> Cc: stable@dpdk.org
>
> Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
> ---
> There are already questions about that "stats" variable:
> - Why a global variable defined in header file?
> - It seems it is not used at all, remove it?
> Forgotten to be removed in below commit, but because of generic
> "stats" name, it didn't cause any compile error but happily used
> global variable
> Fixes: 3810ae435783 ("eventdev: add interrupt driven queues to Rx adapter")
>
> But this patch mainly sent as an RFC, to ask what would you think about
> a cleanup patch to add the library prefix for all the global variables
> that are defined in libries? (yes we have lots of them)
>
> Like:
> zip_cc --> octeontx_zip_cc [1]
> valid_args --> qede_valid_args [2]
> scheduler --> crypto_scheduler [3]
The cleanup patch has been sent:
"fix static variables"
https://patches.dpdk.org/patch/46006/
This patch has been superseded by above one.
@@ -1125,7 +1125,7 @@ rxa_poll(struct rte_event_eth_rx_adapter *rx_adapter)
wrr_pos = rx_adapter->wrr_pos;
max_nb_rx = rx_adapter->max_nb_rx;
buf = &rx_adapter->event_enqueue_buffer;
- stats = &rx_adapter->stats;
+ rte_eventdev_stats = &rx_adapter->stats;
/* Iterate through a WRR sequence */
for (num_queue = 0; num_queue < rx_adapter->wrr_len; num_queue++) {
@@ -592,7 +592,7 @@ typedef int (*eventdev_eth_rx_adapter_stop_t)
(const struct rte_eventdev *dev,
const struct rte_eth_dev *eth_dev);
-struct rte_event_eth_rx_adapter_stats *stats;
+struct rte_event_eth_rx_adapter_stats *rte_eventdev_stats;
/**
* Retrieve ethernet Rx adapter statistics.