| Message ID | 20240903144354.2005690-1-lizetao1@huawei.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [01/21] mtip32xx: Remove redundant null pointer checks in mtip_hw_debugfs_init() | expand |
On 9/3/24 8:43 AM, Li Zetao wrote: > Since the debugfs_create_dir() never returns a null pointer, checking > the return value for a null pointer is redundant, and using IS_ERR is > safe enough. Sigh, why are we seeing so many odd variants of this recently. If you'd do a bit of searching upfront, you'd find that these should not be checked at all rather than changing it from err+null to just an error pointer check. So no to this one, please do at least a tiny bit of research first before blindly making a change based on what some static analyzer told you.
Hi, 在 2024/9/3 22:47, Jens Axboe 写道: > On 9/3/24 8:43 AM, Li Zetao wrote: >> Since the debugfs_create_dir() never returns a null pointer, checking >> the return value for a null pointer is redundant, and using IS_ERR is >> safe enough. > > Sigh, why are we seeing so many odd variants of this recently. If you'd > do a bit of searching upfront, you'd find that these should not be > checked at all rather than changing it from err+null to just an error > pointer check. > > So no to this one, please do at least a tiny bit of research first > before blindly making a change based on what some static analyzer told > you. > I have researched in the community before making the modification. debugfs_create_file can handle illegal dentry, but my understanding is that verifying debugfs_create_dir when it fails can avoid invalid calls to debugfs_create_file. Greg suggested that I remove this check, maybe I can modify it in v2? Thanks, Li Zetao.
On 9/3/24 7:18 PM, Li Zetao wrote: > Hi, > > ? 2024/9/3 22:47, Jens Axboe ??: >> On 9/3/24 8:43 AM, Li Zetao wrote: >>> Since the debugfs_create_dir() never returns a null pointer, checking >>> the return value for a null pointer is redundant, and using IS_ERR is >>> safe enough. >> >> Sigh, why are we seeing so many odd variants of this recently. If you'd >> do a bit of searching upfront, you'd find that these should not be >> checked at all rather than changing it from err+null to just an error >> pointer check. >> >> So no to this one, please do at least a tiny bit of research first >> before blindly making a change based on what some static analyzer told >> you. >> > I have researched in the community before making the modification. > debugfs_create_file can handle illegal dentry, but my understanding is > that verifying debugfs_create_dir when it fails can avoid invalid > calls to debugfs_create_file. > > Greg suggested that I remove this check, maybe I can modify it in v2? debugfs should tolerate error pointers, hence all the error checking should just go away rather than being modified. Something ala the below, totally untested. diff --git a/drivers/block/mtip32xx/mtip32xx.c b/drivers/block/mtip32xx/mtip32xx.c index c6ef0546ffc9..11901f2812ad 100644 --- a/drivers/block/mtip32xx/mtip32xx.c +++ b/drivers/block/mtip32xx/mtip32xx.c @@ -2269,25 +2269,12 @@ static const struct file_operations mtip_flags_fops = { .llseek = no_llseek, }; -static int mtip_hw_debugfs_init(struct driver_data *dd) +static void mtip_hw_debugfs_init(struct driver_data *dd) { - if (!dfs_parent) - return -1; - dd->dfs_node = debugfs_create_dir(dd->disk->disk_name, dfs_parent); - if (IS_ERR_OR_NULL(dd->dfs_node)) { - dev_warn(&dd->pdev->dev, - "Error creating node %s under debugfs\n", - dd->disk->disk_name); - dd->dfs_node = NULL; - return -1; - } - debugfs_create_file("flags", 0444, dd->dfs_node, dd, &mtip_flags_fops); debugfs_create_file("registers", 0444, dd->dfs_node, dd, &mtip_regs_fops); - - return 0; } static void mtip_hw_debugfs_exit(struct driver_data *dd) @@ -4043,10 +4030,6 @@ static int __init mtip_init(void) mtip_major = error; dfs_parent = debugfs_create_dir("rssd", NULL); - if (IS_ERR_OR_NULL(dfs_parent)) { - pr_warn("Error creating debugfs parent\n"); - dfs_parent = NULL; - } /* Register our PCI operations. */ error = pci_register_driver(&mtip_pci_driver);
diff --git a/drivers/block/mtip32xx/mtip32xx.c b/drivers/block/mtip32xx/mtip32xx.c index c6ef0546ffc9..56f3bb30165e 100644 --- a/drivers/block/mtip32xx/mtip32xx.c +++ b/drivers/block/mtip32xx/mtip32xx.c @@ -2275,7 +2275,7 @@ static int mtip_hw_debugfs_init(struct driver_data *dd) return -1; dd->dfs_node = debugfs_create_dir(dd->disk->disk_name, dfs_parent); - if (IS_ERR_OR_NULL(dd->dfs_node)) { + if (IS_ERR(dd->dfs_node)) { dev_warn(&dd->pdev->dev, "Error creating node %s under debugfs\n", dd->disk->disk_name); @@ -4043,7 +4043,7 @@ static int __init mtip_init(void) mtip_major = error; dfs_parent = debugfs_create_dir("rssd", NULL); - if (IS_ERR_OR_NULL(dfs_parent)) { + if (IS_ERR(dfs_parent)) { pr_warn("Error creating debugfs parent\n"); dfs_parent = NULL; }
Since the debugfs_create_dir() never returns a null pointer, checking the return value for a null pointer is redundant, and using IS_ERR is safe enough. Signed-off-by: Li Zetao <lizetao1@huawei.com> --- drivers/block/mtip32xx/mtip32xx.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)