The igen6_edac driver fails to release reference for non-primary memory controllers during both error handling and normal shutdown. After device_initialize() is called in igen6_register_mci(), missing put_device() calls in error paths and igen6_unregister_mcis() cause reference count leaks, resulting in memory leaks and improper device cleanup.
Found by code review.
Cc: stable@vger.kernel.org Fixes: 10590a9d4f23 ("EDAC/igen6: Add EDAC driver for Intel client SoCs using IBECC") Signed-off-by: Ma Ke make24@iscas.ac.cn --- drivers/edac/igen6_edac.c | 4 ++++ 1 file changed, 4 insertions(+)
diff --git a/drivers/edac/igen6_edac.c b/drivers/edac/igen6_edac.c index 2fc59f9eed69..ee2f00b204bb 100644 --- a/drivers/edac/igen6_edac.c +++ b/drivers/edac/igen6_edac.c @@ -1300,6 +1300,8 @@ static int igen6_register_mci(int mc, void __iomem *window, struct pci_dev *pdev imc->mci = mci; return 0; fail3: + if (mc != 0) + put_device(&imc->dev); mci->pvt_info = NULL; kfree(mci->ctl_name); fail2: @@ -1326,6 +1328,8 @@ static void igen6_unregister_mcis(void) kfree(mci->ctl_name); mci->pvt_info = NULL; edac_mc_free(mci); + if (imc->mc != 0) + put_device(&imc->dev); iounmap(imc->window); } }
Hi Ma Ke,
From: Ma Ke make24@iscas.ac.cn Sent: Monday, November 3, 2025 8:40 PM To: Luck, Tony tony.luck@intel.com; Zhuo, Qiuxu qiuxu.zhuo@intel.com; bp@alien8.de Cc: linux-edac@vger.kernel.org; linux-kernel@vger.kernel.org; akpm@linux- foundation.org; Ma Ke make24@iscas.ac.cn; stable@vger.kernel.org Subject: [PATCH] EDAC/igen6: Fix error handling in igen6_edac driver
The igen6_edac driver fails to release reference for non-primary memory controllers during both error handling and normal shutdown. After device_initialize() is called in igen6_register_mci(), missing put_device() calls in error paths and igen6_unregister_mcis() cause reference count leaks, resulting in memory leaks and improper device cleanup.
If igen6_register_mci()fails, kfree(igen6_pvt) is called, and then imc->dev is eventually released. So there are no memory leaks, and put_device() is not needed.
But from the perspective of pairing with device_initialize() for better code readability, then we may add put_device().
Found by code review.
Cc: stable@vger.kernel.org Fixes: 10590a9d4f23 ("EDAC/igen6: Add EDAC driver for Intel client SoCs using
This is NOT a real bug. Please don't add this "Fixes" tag in v2 to avoid unnecessary backporting.
IBECC") Signed-off-by: Ma Ke make24@iscas.ac.cn
drivers/edac/igen6_edac.c | 4 ++++ 1 file changed, 4 insertions(+)
diff --git a/drivers/edac/igen6_edac.c b/drivers/edac/igen6_edac.c index 2fc59f9eed69..ee2f00b204bb 100644 --- a/drivers/edac/igen6_edac.c +++ b/drivers/edac/igen6_edac.c @@ -1300,6 +1300,8 @@ static int igen6_register_mci(int mc, void __iomem *window, struct pci_dev *pdev imc->mci = mci; return 0; fail3:
- if (mc != 0)
Since all memory controllers call device_initialize(), please remove "if (mc != 0)" for unconditionally calls put_device().
put_device(&imc->dev);fail2: @@ -1326,6 +1328,8 @@ static void igen6_unregister_mcis(void) kfree(mci->ctl_name); mci->pvt_info = NULL; edac_mc_free(mci);
if (imc->mc != 0)
Same as above, remove this "if()" check.
put_device(&imc->dev);}
Could you please address the comments above, update the commit messages, and send v2?
Thanks! -Qiuxu
linux-stable-mirror@lists.linaro.org
-
Ma Ke -
Zhuo, Qiuxu