* [PATCH 2/2] ubi: Implement ioctl for detailed erase counters
2024-11-25 13:48 [PATCH 1/2] ubi: Expose interface for detailed erase counters Rickard Andersson
@ 2024-11-25 13:48 ` Rickard Andersson
2024-11-25 15:44 ` [PATCH 1/2] ubi: Expose interface " Richard Weinberger
2024-11-26 5:09 ` Zhihao Cheng
2 siblings, 0 replies; 5+ messages in thread
From: Rickard Andersson @ 2024-11-25 13:48 UTC (permalink / raw)
To: richard, chengzhihao1, linux-mtd, rickard314.andersson
Cc: rickard.andersson, kernel
Currently, only "max_ec" can be read from sysfs, which provides a
limited view of the flash device’s wear. In certain cases, such as
bugs in the wear-leveling algorithm, specific blocks can be worn down
more than others, resulting in uneven wear distribution. Providing
detailed erase counter values give a better understanding of the
overall flash wear.
There exists more detailed info in debugfs, but this information is
only available for debug builds.
Signed-off-by: Rickard Andersson <rickard.andersson@axis.com>
---
drivers/mtd/ubi/cdev.c | 86 ++++++++++++++++++++++++++++++++++++++++++
1 file changed, 86 insertions(+)
diff --git a/drivers/mtd/ubi/cdev.c b/drivers/mtd/ubi/cdev.c
index 0d8f04cf03c5..4af99ddbba66 100644
--- a/drivers/mtd/ubi/cdev.c
+++ b/drivers/mtd/ubi/cdev.c
@@ -828,6 +828,86 @@ static int rename_volumes(struct ubi_device *ubi,
return err;
}
+
+static int ubi_get_ec_info(struct ubi_device *ubi, void __user *argp)
+{
+ struct ubi_ecinfo_req *req;
+ struct ubi_wl_entry *wl;
+ int peb;
+ int end_peb;
+ int i;
+ int err = 0;
+
+ req = kzalloc(sizeof(struct ubi_ecinfo_req), GFP_KERNEL);
+ if (!req)
+ return -ENOMEM;
+
+ /* Copy just the input arguments */
+ err = copy_from_user(req, argp, sizeof(struct ubi_ecinfo_req) -
+ offsetof(struct ubi_ecinfo_req, end_ec_res));
+ if (err) {
+ err = -EFAULT;
+ goto out_free;
+ }
+
+ /* Make sure that start value is less or equal to end value */
+ if (req->start_ec_req > req->end_ec_req) {
+ err = -EINVAL;
+ goto out_free;
+ }
+
+ /* Check input argument start value */
+ if ((req->start_ec_req < 0 || req->start_ec_req >= ubi->peb_count)) {
+ err = -EINVAL;
+ goto out_free;
+ }
+
+ /* Check if end value is less than 0 */
+ if (req->end_ec_req < 0) {
+ err = -EINVAL;
+ goto out_free;
+ }
+
+ end_peb = req->end_ec_req;
+
+ /* Control end value */
+ if (end_peb > ubi->peb_count)
+ end_peb = ubi->peb_count;
+
+ /* Check that we do not overwrite our working memory */
+ if (end_peb > (ARRAY_SIZE(req->erase_counters) + req->start_ec_req))
+ end_peb = ARRAY_SIZE(req->erase_counters) + req->start_ec_req;
+
+ i = 0;
+ for (peb = req->start_ec_req; peb < end_peb; i++, peb++) {
+
+ if (ubi_io_is_bad(ubi, peb)) {
+ req->erase_counters[i] = UBI_UNKNOWN;
+ continue;
+ }
+
+ spin_lock(&ubi->wl_lock);
+
+ wl = ubi->lookuptbl[peb];
+ if (wl)
+ req->erase_counters[i] = wl->ec;
+ else
+ req->erase_counters[i] = UBI_UNKNOWN;
+
+ spin_unlock(&ubi->wl_lock);
+ }
+
+ /* Return last read peb */
+ req->end_ec_res = end_peb - 1;
+
+ if (copy_to_user(argp, req, sizeof(struct ubi_ecinfo_req)))
+ err = -EFAULT;
+
+out_free:
+ kfree(req);
+ return err;
+}
+
static long ubi_cdev_ioctl(struct file *file, unsigned int cmd,
unsigned long arg)
{
@@ -991,6 +1071,12 @@ static long ubi_cdev_ioctl(struct file *file, unsigned int cmd,
break;
}
+ case UBI_IOCECNFO:
+ {
+ err = ubi_get_ec_info(ubi, argp);
+ break;
+ }
+
default:
err = -ENOTTY;
break;
--
2.30.2
______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/
^ permalink raw reply related [flat|nested] 5+ messages in thread* Re: [PATCH 1/2] ubi: Expose interface for detailed erase counters
2024-11-25 13:48 [PATCH 1/2] ubi: Expose interface for detailed erase counters Rickard Andersson
2024-11-25 13:48 ` [PATCH 2/2] ubi: Implement ioctl " Rickard Andersson
@ 2024-11-25 15:44 ` Richard Weinberger
2024-11-25 15:45 ` Richard Weinberger
2024-11-26 5:09 ` Zhihao Cheng
2 siblings, 1 reply; 5+ messages in thread
From: Richard Weinberger @ 2024-11-25 15:44 UTC (permalink / raw)
To: Rickard X Andersson; +Cc: chengzhihao1, linux-mtd, rickard314 andersson, kernel
----- Ursprüngliche Mail -----
> Von: "Rickard X Andersson" <rickard.andersson@axis.com>
> +/**
> + * struct ubi_ecinfo_req - a data structure used in UBI device erase count
> + * information requests and responses.
> + *
> + * @start_ec_req: first erase counter block in range to read
> + * @end_ec_req: last erase counter block in the half-open range
> + * @end_ec_res: last erase counter block that was actually read
> + * @erase_counters: array of erase counter values
> + * @padding: reserved for future, not used, has to be zeroed
> + *
> + * Erase counters are read in range @start_ec_req to @end_ec_req. Range is
> half-open
> + * i.e erase counter of end_ec_req is not actually included.
> + * The read erase counters are placed in @erase_counters. If a block is bad or
> if the
> + * erase counter is unknown the value of the block will be set to -1.
> + */
> +struct ubi_ecinfo_req {
> + __s32 start_ec_req;
> + __s32 end_ec_req;
> + __s32 end_ec_res;
> + __s32 erase_counters[1000];
There is no need to have a limited number. Just use a flexible array.
If userspace asks for more entries than the provided buffer can hold, __put_user()
will fail anyways.
Have a look at fibmap, I think you can the model the API the same way.
Thanks,
//richard
______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/
^ permalink raw reply [flat|nested] 5+ messages in thread* Re: [PATCH 1/2] ubi: Expose interface for detailed erase counters
2024-11-25 15:44 ` [PATCH 1/2] ubi: Expose interface " Richard Weinberger
@ 2024-11-25 15:45 ` Richard Weinberger
0 siblings, 0 replies; 5+ messages in thread
From: Richard Weinberger @ 2024-11-25 15:45 UTC (permalink / raw)
To: Rickard X Andersson; +Cc: chengzhihao1, linux-mtd, rickard314 andersson, kernel
----- Ursprüngliche Mail -----
> Von: "richard" <richard@nod.at>
> An: "Rickard X Andersson" <rickard.andersson@axis.com>
> CC: "chengzhihao1" <chengzhihao1@huawei.com>, "linux-mtd" <linux-mtd@lists.infradead.org>, "rickard314 andersson"
> <rickard314.andersson@gmail.com>, "kernel" <kernel@axis.com>
> Gesendet: Montag, 25. November 2024 16:44:54
> Betreff: Re: [PATCH 1/2] ubi: Expose interface for detailed erase counters
> ----- Ursprüngliche Mail -----
>> Von: "Rickard X Andersson" <rickard.andersson@axis.com>
>> +/**
>> + * struct ubi_ecinfo_req - a data structure used in UBI device erase count
>> + * information requests and responses.
>> + *
>> + * @start_ec_req: first erase counter block in range to read
>> + * @end_ec_req: last erase counter block in the half-open range
>> + * @end_ec_res: last erase counter block that was actually read
>> + * @erase_counters: array of erase counter values
>> + * @padding: reserved for future, not used, has to be zeroed
>> + *
>> + * Erase counters are read in range @start_ec_req to @end_ec_req. Range is
>> half-open
>> + * i.e erase counter of end_ec_req is not actually included.
>> + * The read erase counters are placed in @erase_counters. If a block is bad or
>> if the
>> + * erase counter is unknown the value of the block will be set to -1.
>> + */
>> +struct ubi_ecinfo_req {
>> + __s32 start_ec_req;
>> + __s32 end_ec_req;
>> + __s32 end_ec_res;
>> + __s32 erase_counters[1000];
>
> There is no need to have a limited number. Just use a flexible array.
> If userspace asks for more entries than the provided buffer can hold,
> __put_user()
> will fail anyways.
>
> Have a look at fibmap, I think you can the model the API the same way.
*fiemap
Thanks,
//richard
______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH 1/2] ubi: Expose interface for detailed erase counters
2024-11-25 13:48 [PATCH 1/2] ubi: Expose interface for detailed erase counters Rickard Andersson
2024-11-25 13:48 ` [PATCH 2/2] ubi: Implement ioctl " Rickard Andersson
2024-11-25 15:44 ` [PATCH 1/2] ubi: Expose interface " Richard Weinberger
@ 2024-11-26 5:09 ` Zhihao Cheng
2 siblings, 0 replies; 5+ messages in thread
From: Zhihao Cheng @ 2024-11-26 5:09 UTC (permalink / raw)
To: Rickard Andersson, richard, linux-mtd, rickard314.andersson; +Cc: kernel
在 2024/11/25 21:48, Rickard Andersson 写道:
> Using the ioctl command 'UBI_IOCECNFO' user space can obtain
> detailed erase counter information of all blocks of a device.
>
> Signed-off-by: Rickard Andersson <rickard.andersson@axis.com>
> ---
> include/uapi/mtd/ubi-user.h | 25 +++++++++++++++++++++++++
> 1 file changed, 25 insertions(+)
>
> diff --git a/include/uapi/mtd/ubi-user.h b/include/uapi/mtd/ubi-user.h
> index e1571603175e..667680e44ed4 100644
> --- a/include/uapi/mtd/ubi-user.h
> +++ b/include/uapi/mtd/ubi-user.h
> @@ -175,6 +175,8 @@
> #define UBI_IOCRPEB _IOW(UBI_IOC_MAGIC, 4, __s32)
> /* Force scrubbing on the specified PEB */
> #define UBI_IOCSPEB _IOW(UBI_IOC_MAGIC, 5, __s32)
> +/* Read detailed device erase counter information */
> +#define UBI_IOCECNFO _IOWR(UBI_IOC_MAGIC, 6, struct ubi_ecinfo_req)
>
> /* ioctl commands of the UBI control character device */
>
> @@ -470,4 +472,27 @@ struct ubi_blkcreate_req {
> __s8 padding[128];
> } __packed;
>
> +/**
> + * struct ubi_ecinfo_req - a data structure used in UBI device erase count
> + * information requests and responses.
> + *
> + * @start_ec_req: first erase counter block in range to read
> + * @end_ec_req: last erase counter block in the half-open range
> + * @end_ec_res: last erase counter block that was actually read
> + * @erase_counters: array of erase counter values
> + * @padding: reserved for future, not used, has to be zeroed
> + *
> + * Erase counters are read in range @start_ec_req to @end_ec_req. Range is half-open
> + * i.e erase counter of end_ec_req is not actually included.
> + * The read erase counters are placed in @erase_counters. If a block is bad or if the
> + * erase counter is unknown the value of the block will be set to -1.
> + */
> +struct ubi_ecinfo_req {
> + __s32 start_ec_req;
> + __s32 end_ec_req;
> + __s32 end_ec_res;
> + __s32 erase_counters[1000];
> + __s8 padding[84];
> +} __packed;
> +
Hi Rickard, please move the definition of 'ubi_ecinfo_req' between
'ubi_rnvol_req' and 'ubi_leb_change_req'. The last part in
'include/uapi/mtd/ubi-user.h' defines all structures for the ubi volume
ioctl.
> #endif /* __UBI_USER_H__ */
>
______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/
^ permalink raw reply [flat|nested] 5+ messages in thread