| Message ID | 20230926083229.2073890-1-u.kleine-koenig@pengutronix.de (mailing list archive) |
|---|---|
| State | Accepted |
| Commit | f1a1bc8775b26345aba2be278118999e7f661d3d |
| Headers | show |
| Series | soc: qcom: llcc: Handle a second device without data corruption | expand |
Hello, On Tue, Sep 26, 2023 at 10:32:29AM +0200, Uwe Kleine-König wrote: > the check added here is racy. To fix that race a lock would be needed to > protect drv_data. Alternatively (and cleaner) drv_data could be stored > using platform_set_drvdata(pdev, ...), but that cannot be done trivially > because there is a global function (llcc_slice_getd) that needs drv_data > but has no access to the platform device. > > I don't know what is sensible here, so I just added this 90% fix. If you > want a more sophisticated solution, please create one yourself and just > consider my patch a problem report. if you consider this a fix, the corresponding tag line is: Fixes: a3134fb09e0b ("drivers: soc: Add LLCC driver") Best regards Uwe
On Tue, 26 Sep 2023 10:32:29 +0200, Uwe Kleine-König wrote: > Usually there is only one llcc device. But if there were a second, even > a failed probe call would modify the global drv_data pointer. So check > if drv_data is valid before overwriting it. > > Applied, thanks! [1/1] soc: qcom: llcc: Handle a second device without data corruption commit: f1a1bc8775b26345aba2be278118999e7f661d3d Best regards,
diff --git a/drivers/soc/qcom/llcc-qcom.c b/drivers/soc/qcom/llcc-qcom.c index feb21637ac20..e3f262a50e4a 100644 --- a/drivers/soc/qcom/llcc-qcom.c +++ b/drivers/soc/qcom/llcc-qcom.c @@ -1120,6 +1120,9 @@ static int qcom_llcc_probe(struct platform_device *pdev) u32 version; struct regmap *regmap; + if (!IS_ERR(drv_data)) + return -EBUSY; + drv_data = devm_kzalloc(dev, sizeof(*drv_data), GFP_KERNEL); if (!drv_data) { ret = -ENOMEM;
Usually there is only one llcc device. But if there were a second, even a failed probe call would modify the global drv_data pointer. So check if drv_data is valid before overwriting it. Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> --- Hello, the check added here is racy. To fix that race a lock would be needed to protect drv_data. Alternatively (and cleaner) drv_data could be stored using platform_set_drvdata(pdev, ...), but that cannot be done trivially because there is a global function (llcc_slice_getd) that needs drv_data but has no access to the platform device. I don't know what is sensible here, so I just added this 90% fix. If you want a more sophisticated solution, please create one yourself and just consider my patch a problem report. Best regards Uwe drivers/soc/qcom/llcc-qcom.c | 3 +++ 1 file changed, 3 insertions(+)