Re: [PATCH v3 1/2] tools/counter: add a flexible watch events tool

From: Fabrice Gasnier
Date: Mon Dec 11 2023 - 12:06:43 EST


On 12/11/23 16:57, William Breathitt Gray wrote:
> On Wed, Dec 06, 2023 at 05:47:25PM +0100, Fabrice Gasnier wrote:
>> This adds a new counter tool to be able to test various watch events.
>> A flexible watch array can be populated from command line, each field
>> may be tuned with a dedicated command line sub-option in "--watch" string.
>> Several watch events can be defined, each can have specific watch options,
>> by using "--watch <watch 1 options> --watch <watch 2 options>".
>> Watch options is a comma separated list.
>>
>> It also comes with a simple default watch (to monitor overflow/underflow
>> events), used when no watch parameters are provided. It's equivalent to:
>> counter_watch_events -w comp_count,scope_count,evt_ovf_udf
>>
>> The print_usage() routine proposes another example, from the command line,
>> which generates a 2 elements watch array, to monitor:
>> - overflow underflow events
>> - capture events, on channel 3, that reads read captured data by
>> specifying the component id (capture3_component_id being 7 here).
>>
>> Signed-off-by: Fabrice Gasnier <fabrice.gasnier@xxxxxxxxxxx>
>> ---
>> Changes in v3:
>> - Free the allocated memory, also close the char device
>> - Split of another patch series[1].
>> [1] https://lore.kernel.org/lkml/20230922143920.3144249-1-fabrice.gasnier@xxxxxxxxxxx/
>
> Hi Fabrice,
>
> Thank you for splitting this from the other patches. I think you may
> still be leaking memory in a few places below.
>
>> + if (nwatch) {
>> + watches = calloc(nwatch, sizeof(*watches));
>> + if (!watches) {
>> + perror("Error allocating watches");
>> + return 1;
>> + }
>> + } else {
>> + /* default to simple watch example */
>> + watches = simple_watch;
>> + nwatch = ARRAY_SIZE(simple_watch);
>> + }
>
> If we go down the calloc() path, then we should free the memory
> before any return.

Hi William,

Ah yes, I missed that. I'll fix it in later revision, Thanks!

Best Regards,
Fabrice

>
>> + case WATCH_CHANNEL:
>> + if (!value) {
>> + fprintf(stderr, "Missing chan=<number>\n");
>> + return -EINVAL;
>
> Such as here.
>
>> + }
>> + watches[i].channel = strtoul(value, NULL, 10);
>> + if (errno)
>> + return -errno;
>
> Here.
>
>> + break;
>> + case WATCH_ID:
>> + if (!value) {
>> + fprintf(stderr, "Missing id=<number>\n");
>> + return -EINVAL;
>
> Here.
>
>> + }
>> + watches[i].component.id = strtoul(value, NULL, 10);
>> + if (errno)
>> + return -errno;
>
> Here.
>
>> + break;
>> + case WATCH_PARENT:
>> + if (!value) {
>> + fprintf(stderr, "Missing parent=<number>\n");
>> + return -EINVAL;
>
> Here.
>
>> + }
>> + watches[i].component.parent = strtoul(value, NULL, 10);
>> + if (errno)
>> + return -errno;
>
> Here.
>
>> + break;
>> + default:
>> + fprintf(stderr, "Unknown suboption '%s'\n", value);
>> + return -EINVAL;
>
> Here.
>
>> + ret = asprintf(&device_name, "/dev/counter%d", dev_num);
>> + if (ret < 0)
>> + return -ENOMEM;
>
> Here.
>
>> + fd = open(device_name, O_RDWR);
>> + if (fd == -1) {
>> + perror("Unable to open counter device");
>> + return 1;
>
> Here.
>
>> + }
>> +
>> + for (i = 0; i < nwatch; i++) {
>> + ret = ioctl(fd, COUNTER_ADD_WATCH_IOCTL, watches + i);
>> + if (ret == -1) {
>> + fprintf(stderr, "Error adding watches[%d]: %s\n", i,
>> + strerror(errno));
>> + return 1;
>
> Here.
>
>> + }
>> + }
>> +
>> + ret = ioctl(fd, COUNTER_ENABLE_EVENTS_IOCTL);
>> + if (ret == -1) {
>> + perror("Error enabling events");
>> + return 1;
>
> Here.
>
>> + }
>> +
>> + for (i = 0; loop <= 0 || i < loop; i++) {
>> + ret = read(fd, &event_data, sizeof(event_data));
>> + if (ret == -1) {
>> + perror("Failed to read event data");
>> + return 1;
>
> Here.
>
>> + }
>> +
>> + if (ret != sizeof(event_data)) {
>> + fprintf(stderr, "Failed to read event data\n");
>> + return -EIO;
>
> And here.
>
>> + if (watches != simple_watch)
>> + free(watches);
>> + close(fd);
>> +
>> + return 0;
>
> We finally free watches here, close the file descriptor, and return. So
> instead of returning an error code directly when you encounter a
> problem, I would do an unwinding goto section like the following
> instead.
>
> First, the open() call occurs after the calloc(), so move the close()
> call above the watches check so that we unwind in a first-in-last-out
> order. Next, add a label to mark the file descriptor close section, and
> another label to mark the watches free section. Then, rather than
> returning 0 directly, return a retval that we can set. That way, when
> you need to return on an error, set retval to the error code and goto
> the file descriptor close label if we're past the open() call, or the
> watches free label if we're just past the calloc() call.
>
> William Breathitt Gray