| Message ID | 20230815125233.65469-1-lihuisong@huawei.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | soc: kunpeng_hccs: fix some sparse warnings about incorrect type | expand |
On Tue, Aug 15, 2023, at 14:52, Huisong Li wrote: > @@ -199,8 +200,8 @@ static int hccs_pcc_cmd_send(struct hccs_dev *hdev, u8 cmd, > struct hccs_desc *desc) > { > struct hccs_mbox_client_info *cl_info = &hdev->cl_info; > - struct acpi_pcct_shared_memory *comm_base = cl_info->pcc_comm_addr; > - void *comm_space = (void *)(comm_base + 1); > + void __iomem *comm_space = (u8 *)cl_info->pcc_comm_addr + > + sizeof(struct acpi_pcct_shared_memory); > struct hccs_fw_inner_head *fw_inner_head; > struct acpi_pcct_shared_memory tmp = {0}; > u16 comm_space_size; The cast still looks wrong, while both comm_space and ->pcc_comm_addr are __iomem pointers, casting to a non-__iomem u8 pointer should still produce that warning. I'd suggest casting to (u8 __iomem *) or (void __iomem *), I assume you still need a cast here to get the correct address. Arnd
在 2023/8/16 21:17, Arnd Bergmann 写道: > On Tue, Aug 15, 2023, at 14:52, Huisong Li wrote: >> @@ -199,8 +200,8 @@ static int hccs_pcc_cmd_send(struct hccs_dev *hdev, u8 cmd, >> struct hccs_desc *desc) >> { >> struct hccs_mbox_client_info *cl_info = &hdev->cl_info; >> - struct acpi_pcct_shared_memory *comm_base = cl_info->pcc_comm_addr; >> - void *comm_space = (void *)(comm_base + 1); >> + void __iomem *comm_space = (u8 *)cl_info->pcc_comm_addr + >> + sizeof(struct acpi_pcct_shared_memory); >> struct hccs_fw_inner_head *fw_inner_head; >> struct acpi_pcct_shared_memory tmp = {0}; >> u16 comm_space_size; > The cast still looks wrong, while both comm_space and ->pcc_comm_addr > are __iomem pointers, casting to a non-__iomem u8 pointer should > still produce that warning. I'd suggest casting to (u8 __iomem *) > or (void __iomem *), I assume you still need a cast here to get > the correct address. Thanks Arnd. You are right. So I modify it by dropping (u8 *) here in patch v2. Please take a look at this again. If it is ok for you, can you apply this patch to soc/soc.git? /Huisong
On Thu, Aug 17, 2023, at 03:37, lihuisong (C) wrote: > 在 2023/8/16 21:17, Arnd Bergmann 写道: >> On Tue, Aug 15, 2023, at 14:52, Huisong Li wrote: >>> @@ -199,8 +200,8 @@ static int hccs_pcc_cmd_send(struct hccs_dev *hdev, u8 cmd, >>> struct hccs_desc *desc) >>> { >>> struct hccs_mbox_client_info *cl_info = &hdev->cl_info; >>> - struct acpi_pcct_shared_memory *comm_base = cl_info->pcc_comm_addr; >>> - void *comm_space = (void *)(comm_base + 1); >>> + void __iomem *comm_space = (u8 *)cl_info->pcc_comm_addr + >>> + sizeof(struct acpi_pcct_shared_memory); >>> struct hccs_fw_inner_head *fw_inner_head; >>> struct acpi_pcct_shared_memory tmp = {0}; >>> u16 comm_space_size; >> The cast still looks wrong, while both comm_space and ->pcc_comm_addr >> are __iomem pointers, casting to a non-__iomem u8 pointer should >> still produce that warning. I'd suggest casting to (u8 __iomem *) >> or (void __iomem *), I assume you still need a cast here to get >> the correct address. > Thanks Arnd. > > You are right. So I modify it by dropping (u8 *) here in patch v2. > Please take a look at this again. Right, if pcc_comm_addr is already a u8 __iomem* or void*, that works, I though the cast was needed since it points to a larger object, which would throw off the calculation. > If it is ok for you, can you apply this patch to soc/soc.git? Yes, of course. Arnd
在 2023/8/17 19:54, Arnd Bergmann 写道: > On Thu, Aug 17, 2023, at 03:37, lihuisong (C) wrote: >> 在 2023/8/16 21:17, Arnd Bergmann 写道: >>> On Tue, Aug 15, 2023, at 14:52, Huisong Li wrote: >>>> @@ -199,8 +200,8 @@ static int hccs_pcc_cmd_send(struct hccs_dev *hdev, u8 cmd, >>>> struct hccs_desc *desc) >>>> { >>>> struct hccs_mbox_client_info *cl_info = &hdev->cl_info; >>>> - struct acpi_pcct_shared_memory *comm_base = cl_info->pcc_comm_addr; >>>> - void *comm_space = (void *)(comm_base + 1); >>>> + void __iomem *comm_space = (u8 *)cl_info->pcc_comm_addr + >>>> + sizeof(struct acpi_pcct_shared_memory); >>>> struct hccs_fw_inner_head *fw_inner_head; >>>> struct acpi_pcct_shared_memory tmp = {0}; >>>> u16 comm_space_size; >>> The cast still looks wrong, while both comm_space and ->pcc_comm_addr >>> are __iomem pointers, casting to a non-__iomem u8 pointer should >>> still produce that warning. I'd suggest casting to (u8 __iomem *) >>> or (void __iomem *), I assume you still need a cast here to get >>> the correct address. >> Thanks Arnd. >> >> You are right. So I modify it by dropping (u8 *) here in patch v2. >> Please take a look at this again. > Right, if pcc_comm_addr is already a u8 __iomem* or void*, that > works, I though the cast was needed since it points to a larger > object, which would throw off the calculation. This pcc_comm_addr is already defined "void __iomem *" type in patch v2. So that works well, right? /Huisong > >> If it is ok for you, can you apply this patch to soc/soc.git? > Yes, of course. > > Arnd > .
On Fri, Aug 18, 2023, at 03:38, lihuisong (C) wrote: > 在 2023/8/17 19:54, Arnd Bergmann 写道: >> On Thu, Aug 17, 2023, at 03:37, lihuisong (C) wrote: >>> 在 2023/8/16 21:17, Arnd Bergmann 写道: >>>> On Tue, Aug 15, 2023, at 14:52, Huisong Li wrote: >>>>> @@ -199,8 +200,8 @@ static int hccs_pcc_cmd_send(struct hccs_dev *hdev, u8 cmd, >>>>> struct hccs_desc *desc) >>>>> { >>>>> struct hccs_mbox_client_info *cl_info = &hdev->cl_info; >>>>> - struct acpi_pcct_shared_memory *comm_base = cl_info->pcc_comm_addr; >>>>> - void *comm_space = (void *)(comm_base + 1); >>>>> + void __iomem *comm_space = (u8 *)cl_info->pcc_comm_addr + >>>>> + sizeof(struct acpi_pcct_shared_memory); >>>>> struct hccs_fw_inner_head *fw_inner_head; >>>>> struct acpi_pcct_shared_memory tmp = {0}; >>>>> u16 comm_space_size; >>>> The cast still looks wrong, while both comm_space and ->pcc_comm_addr >>>> are __iomem pointers, casting to a non-__iomem u8 pointer should >>>> still produce that warning. I'd suggest casting to (u8 __iomem *) >>>> or (void __iomem *), I assume you still need a cast here to get >>>> the correct address. >>> Thanks Arnd. >>> >>> You are right. So I modify it by dropping (u8 *) here in patch v2. >>> Please take a look at this again. >> Right, if pcc_comm_addr is already a u8 __iomem* or void*, that >> works, I though the cast was needed since it points to a larger >> object, which would throw off the calculation. > This pcc_comm_addr is already defined "void __iomem *" type in patch v2. > So that works well, right? Yes, that works. Arnd
diff --git a/drivers/soc/hisilicon/kunpeng_hccs.c b/drivers/soc/hisilicon/kunpeng_hccs.c index 0d6f6bacd3f6..4e94753ab0f7 100644 --- a/drivers/soc/hisilicon/kunpeng_hccs.c +++ b/drivers/soc/hisilicon/kunpeng_hccs.c @@ -156,8 +156,8 @@ static int hccs_register_pcc_channel(struct hccs_dev *hdev) } if (pcc_chan->shmem_base_addr) { - cl_info->pcc_comm_addr = (void __force *)ioremap( - pcc_chan->shmem_base_addr, pcc_chan->shmem_size); + cl_info->pcc_comm_addr = ioremap(pcc_chan->shmem_base_addr, + pcc_chan->shmem_size); if (!cl_info->pcc_comm_addr) { dev_err(dev, "Failed to ioremap PCC communication region for channel-%d.\n", hdev->chan_id); @@ -177,7 +177,8 @@ static int hccs_register_pcc_channel(struct hccs_dev *hdev) static int hccs_check_chan_cmd_complete(struct hccs_dev *hdev) { struct hccs_mbox_client_info *cl_info = &hdev->cl_info; - struct acpi_pcct_shared_memory *comm_base = cl_info->pcc_comm_addr; + struct acpi_pcct_shared_memory __iomem *comm_base = + cl_info->pcc_comm_addr; u16 status; int ret; @@ -199,8 +200,8 @@ static int hccs_pcc_cmd_send(struct hccs_dev *hdev, u8 cmd, struct hccs_desc *desc) { struct hccs_mbox_client_info *cl_info = &hdev->cl_info; - struct acpi_pcct_shared_memory *comm_base = cl_info->pcc_comm_addr; - void *comm_space = (void *)(comm_base + 1); + void __iomem *comm_space = (u8 *)cl_info->pcc_comm_addr + + sizeof(struct acpi_pcct_shared_memory); struct hccs_fw_inner_head *fw_inner_head; struct acpi_pcct_shared_memory tmp = {0}; u16 comm_space_size; @@ -212,7 +213,7 @@ static int hccs_pcc_cmd_send(struct hccs_dev *hdev, u8 cmd, tmp.command = cmd; /* Clear cmd complete bit */ tmp.status = 0; - memcpy_toio(comm_base, (void *)&tmp, + memcpy_toio(cl_info->pcc_comm_addr, (void *)&tmp, sizeof(struct acpi_pcct_shared_memory)); /* Copy the message to the PCC comm space */ diff --git a/drivers/soc/hisilicon/kunpeng_hccs.h b/drivers/soc/hisilicon/kunpeng_hccs.h index 9d71fb78443f..6012d2776028 100644 --- a/drivers/soc/hisilicon/kunpeng_hccs.h +++ b/drivers/soc/hisilicon/kunpeng_hccs.h @@ -50,7 +50,7 @@ struct hccs_mbox_client_info { struct mbox_chan *mbox_chan; struct pcc_mbox_chan *pcc_chan; u64 deadline_us; - void *pcc_comm_addr; + void __iomem *pcc_comm_addr; }; struct hccs_dev {
This patch fixes some sparse warnings about incorrect type. The address about PCC communication space should use '__iomem' tag. Reported-by: kernel test robot <lkp@intel.com> Closes: https://lore.kernel.org/oe-kbuild-all/202308151142.dH5Muhva-lkp@intel.com/ Fixes: 886bdf9c883b ("soc: hisilicon: Support HCCS driver on Kunpeng SoC") Signed-off-by: Huisong Li <lihuisong@huawei.com> --- drivers/soc/hisilicon/kunpeng_hccs.c | 13 +++++++------ drivers/soc/hisilicon/kunpeng_hccs.h | 2 +- 2 files changed, 8 insertions(+), 7 deletions(-)