ufs: core: Always read the descriptors with max length
Subsystems (2)
v5 cover ufs: core: Always read the descriptors with max length
v4--v5:
Change patch 2 according to Bart's comment
v3--v4:
Add "Reviewed-by" to patch's commits
Use kzalloc instead of kmalloc in drivers/ufs/core/ufshcd.c - patch 2/4
v2--v3:
Based on Bean's comments:
1)Use kzalloc instead of kmalloc in ufshcd_set_active_icc_lvl - patch 2/4 2)Delete UFS_RPMB_UNIT definition - patch 2/4
3)Delete len description - patch 3/4
v1--v2:
Fix argument warning in ufshpb.c
Read any descriptor with a maximum size of QUERY_DESC_MAX_SIZE. According to the spec the device rerurns the actual size. Thus can improve code readability and save CPU cycles.
While at it, cleanup few leftovers around the descriptor size parameter.
Arthur Simchaev (4):
ufs:core: Remove redundant wb check
ufs:core: Remove redundant desc_size variable from hba
ufs: core: Remove len parameter from ufshcd_set_active_icc_lvl
ufs: core: Remove ufshcd_map_desc_id_to_length function
drivers/ufs/core/ufs_bsg.c | 7 +-- drivers/ufs/core/ufshcd-priv.h | 3 -- drivers/ufs/core/ufshcd.c | 100 ++++++++++------------------------------- drivers/ufs/core/ufshpb.c | 5 +-- include/ufs/ufshcd.h | 1 - 5 files changed, 26 insertions(+), 90 deletions(-)
Signature · 1 line
2.7.4
PATCH v5 1/4 ufs: core: Remove redundant wb check
We used to use the extended-feature field in the device descriptor, as an indication that the device supports ufs2.2 or later. Remove that as this check is specifically done few lines above.
--- drivers/ufs/core/ufshcd.c | 4 ---- 1 file changed, 4 deletions(-)
drivers/ufs/core/ufshcd.c +0 -4
@@ -7608,10 +7608,6 @@ static void ufshcd_wb_probe(struct ufs_hba *hba, const u8 *desc_buf) (hba->dev_quirks & UFS_DEVICE_QUIRK_SUPPORT_EXTENDED_FEATURES))) goto wb_disabled; -if (hba->desc_size[QUERY_DESC_IDN_DEVICE ] < -DEVICE_DESC_PARAM_EXT_UFS_FEATURE_SUP +4 ) - goto wb_disabled; - ext_ufs_feature =get_unaligned_be32 (desc_buf +DEVICE_DESC_PARAM_EXT_UFS_FEATURE_SUP );
Signature · 1 line
2.7.4
Re: [PATCH v5 1/4] ufs: core: Remove redundant wb check
Quoted 8 lines · Arthur Simchaev
We used to use the extended-feature field in the device descriptor, as an indication that the device supports ufs2.2 or later. Remove that as this check is specifically done few lines above.
Reviewed-by Bart Van Assche <[email protected]>Reviewed-by Bean Huo <[email protected]>Signed-off-by Arthur Simchaev <[email protected]>
PATCH v5 2/4 ufs: core: Remove redundant desc_size variable from hba
In reply to Arthur SimchaevAlways read the descriptor with QUERY_DESC_MAX_SIZE. According to the spec, the device returns the actual size
--- drivers/ufs/core/ufs_bsg.c | 6 +--- drivers/ufs/core/ufshcd-priv.h | 3 -- drivers/ufs/core/ufshcd.c | 72 ++++++++---------------------------------- drivers/ufs/core/ufshpb.c | 4 +-- include/ufs/ufs.h | 1 - include/ufs/ufshcd.h | 4 --- 6 files changed, 15 insertions(+), 75 deletions(-)
drivers/ufs/core/ufs_bsg.c +1 -5
@@ -21,11 +21,7 @@ static int ufs_bsg_get_query_desc_size(struct ufs_hba *hba, int *desc_len,if (desc_size <=0 )return -EINVAL ; -ufshcd_map_desc_id_to_length (hba, desc_id, desc_len); -if (!*desc_len) -return -EINVAL ; - - *desc_len =min_t (int, *desc_len, desc_size); + *desc_len =min_t (int,QUERY_DESC_MAX_SIZE , desc_size);return 0 ; }
drivers/ufs/core/ufshcd-priv.h +0 -3
@@ -70,9 +70,6 @@ int ufshcd_read_string_desc(struct ufs_hba *hba, u8 desc_index,int ufshcd_hold (struct ufs_hba *hba,bool async);void ufshcd_release (struct ufs_hba *hba); -void ufshcd_map_desc_id_to_length (struct ufs_hba *hba,enum desc_idn desc_id, -int *desc_length); -int ufshcd_send_uic_cmd (struct ufs_hba *hba,struct uic_command *uic_cmd); int ufshcd_exec_raw_upiu_cmd(struct ufs_hba *hba,
drivers/ufs/core/ufshcd.c +13 -59
159 lines not on this page · raw message
drivers/ufs/core/ufshpb.c +1 -3
@@ -2382,12 +2382,10 @@ static int ufshpb_get_lu_info(struct ufs_hba *hba, int lun, {u16 max_active_rgns;u8 lu_enable; -int size; +int size =QUERY_DESC_MAX_SIZE ;int ret;char desc_buf[QUERY_DESC_MAX_SIZE ]; -ufshcd_map_desc_id_to_length (hba,QUERY_DESC_IDN_UNIT , &size); -ufshcd_rpm_get_sync (hba); ret = ufshcd_query_descriptor_retry(hba, UPIU_QUERY_OPCODE_READ_DESC, QUERY_DESC_IDN_UNIT, lun, 0,
include/ufs/ufs.h +0 -1
@@ -38,7 +38,6 @@#define UFS_UPIU_MAX_UNIT_NUM_ID 0x7F#define UFS_MAX_LUNS (SCSI_W_LUN_BASE + UFS_UPIU_MAX_UNIT_NUM_ID)#define UFS_UPIU_WLUN_ID (1 << 7) -#define UFS_RPMB_UNIT 0xC4/* WriteBooster buffer is available only for the logical unit from 0 to 7 */ #define UFS_UPIU_MAX_WB_LUN_ID 8
include/ufs/ufshcd.h +0 -4
@@ -952,7 +952,6 @@ struct ufs_hba {bool is_urgent_bkops_lvl_checked;struct rw_semaphore clk_scaling_lock; -unsigned char desc_size[QUERY_DESC_IDN_MAX ];atomic_t scsi_block_reqs_cnt;struct device bsg_dev;
@@ -1186,9 +1185,6 @@ void ufshcd_release(struct ufs_hba *hba);void ufshcd_clkgate_delay_set (struct device *dev,unsigned long value); -void ufshcd_map_desc_id_to_length (struct ufs_hba *hba,enum desc_idn desc_id, -int *desc_length); -u32 ufshcd_get_local_unipro_ver (struct ufs_hba *hba);int ufshcd_get_vreg (struct device *dev,struct ufs_vreg *vreg);
Signature · 1 line
2.7.4
Re: [PATCH v5 2/4] ufs: core: Remove redundant desc_size variable from hba
Quoted 2 lines · Arthur Simchaev
Always read the descriptor with
QUERY_DESC_MAX_SIZE. According to the spec, the device returns the actual size
PATCH v5 3/4 ufs: core: Remove len parameter from ufshcd_set_active_icc_lvl
In reply to Arthur Simchaevlen argument is not used anymore in ufshcd_set_active_icc_lvl function.
--- drivers/ufs/core/ufshcd.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-)
drivers/ufs/core/ufshcd.c +2 -4
@@ -7394,12 +7394,11 @@ static u32 ufshcd_get_max_icc_level(int sup_curr_uA, u32 start_scan,* In case regulators are not initialized we'll return 0 * @hba: per-adapter instance * @desc_buf: power descriptor buffer to extract ICC levels from. -* @len: length of desc_buff * * Returns calculated ICC level */ static u32 ufshcd_find_max_sup_active_icc_level (struct ufs_hba *hba, -const u8 *desc_buf,int len) +const u8 *desc_buf) {u32 icc_level =0 ;
@@ -7457,8 +7456,7 @@ static void ufshcd_set_active_icc_lvl(struct ufs_hba *hba) goto out; } - icc_level =ufshcd_find_max_sup_active_icc_level (hba, desc_buf, -QUERY_DESC_MAX_SIZE ); + icc_level =ufshcd_find_max_sup_active_icc_level (hba, desc_buf);dev_dbg (hba->dev,"%s: setting icc_level 0x%x" , __func__, icc_level); ret = ufshcd_query_attr_retry(hba, UPIU_QUERY_OPCODE_WRITE_ATTR,
Signature · 1 line
2.7.4
PATCH v5 4/4 ufs: core: Remove ufshcd_map_desc_id_to_length function
In reply to Arthur SimchaevThere shouldn't be any restriction of the descriptor size (not the descriptor id for that matter) up to QUERY_DESC_MAX_SIZE. According to the spec, the caller can use any descriptor size, and it is up to the device to return the actual size. Therefore there shouldn't be any sizes hardcoded in the kernel, nor any need to cache it, hence ufshcd_map_desc_id_to_length function is redundant. Always read the descriptors with QUERY_DESC_MAX_SIZE size.
--- drivers/ufs/core/ufs_bsg.c | 1 - drivers/ufs/core/ufshcd.c | 23 +++++++++++------------ 2 files changed, 11 insertions(+), 13 deletions(-)
drivers/ufs/core/ufs_bsg.c +0 -1
@@ -16,7 +16,6 @@ static int ufs_bsg_get_query_desc_size(struct ufs_hba *hba, int *desc_len, struct utp_upiu_query *qr) {int desc_size =be16_to_cpu (qr->length); -int desc_id = qr->idn;if (desc_size <=0 )return -EINVAL ;
drivers/ufs/core/ufshcd.c +11 -12
@@ -3395,12 +3395,6 @@ int ufshcd_read_desc_param(struct ufs_hba *hba,if (desc_id >=QUERY_DESC_IDN_MAX || !param_size)return -EINVAL ; -if (param_offset >= buff_len) { -dev_err (hba->dev,"%s: Invalid offset 0x%x in descriptor IDN 0x%x, length 0x%x\n" , - __func__, param_offset, desc_id, buff_len); -return -EINVAL ; - } -/* Check whether we need temp memory */ if (param_offset !=0 || param_size < buff_len) { desc_buf =kzalloc (buff_len,GFP_KERNEL );
@@ -3413,15 +3407,23 @@ int ufshcd_read_desc_param(struct ufs_hba *hba,/* Request for full descriptor */ ret =ufshcd_query_descriptor_retry (hba,UPIU_QUERY_OPCODE_READ_DESC , - desc_id, desc_index,0 , - desc_buf, &buff_len); - + desc_id, desc_index,0 , + desc_buf, &buff_len);if (ret) {dev_err (hba->dev,"%s: Failed reading descriptor. desc_id %d, desc_index %d, param_offset %d, ret %d\n" , __func__, desc_id, desc_index, param_offset, ret); goto out; } +/* Update descriptor length */ + buff_len = desc_buf[QUERY_DESC_LENGTH_OFFSET ]; + +if (param_offset >= buff_len) { +dev_err (hba->dev,"%s: Invalid offset 0x%x in descriptor IDN 0x%x, length 0x%x\n" , + __func__, param_offset, desc_id, buff_len); +return -EINVAL ; + } +/* Sanity check */ if (desc_buf[QUERY_DESC_DESC_TYPE_OFFSET ] != desc_id) { dev_err(hba->dev, "%s: invalid desc_id %d in descriptor header\n",
@@ -3430,9 +3432,6 @@ int ufshcd_read_desc_param(struct ufs_hba *hba, goto out; } -/* Update descriptor length */ - buff_len = desc_buf[QUERY_DESC_LENGTH_OFFSET ]; -if (is_kmalloc) {/* Make sure we don't copy more data than available */ if (param_offset >= buff_len)
Signature · 1 line
2.7.4
Bart Van Assche
reviewerRe: [PATCH v5 4/4] ufs: core: Remove ufshcd_map_desc_id_to_length function
Quoted 9 lines · Arthur Simchaev
There shouldn't be any restriction of the descriptor size (not the descriptor id for that matter) up to
QUERY_DESC_MAX_SIZE. According to the spec, the caller can use any descriptor size, and it is up to the device to return the actual size. Therefore there shouldn't be any sizes hardcoded in the kernel, nor any need to cache it, henceufshcd_map_desc_id_to_lengthfunction is redundant. Always read the descriptors withQUERY_DESC_MAX_SIZEsize.Reviewed-by Bart Van Assche <[email protected]>
I have not yet replied with "Reviewed-by" to this patch so you are not yet allowed to add my Reviewed-by tag to this patch.
drivers/ufs/core/ufshcd.c int ufshcd_read_desc_param(struct ufs_hba *hba, · line 3419
Quoted 2 lines
drivers/ufs/core/ufshcd.c +2
-0+/* Update descriptor length */ + buff_len = desc_buf[QUERY_DESC_LENGTH_OFFSET ];
Should a check be added here that desc_buf[QUERY_DESC_LENGTH_OFFSET] <= QUERY_DESC_MAX_SIZE to protect against firmware bugs? Since QUERY_DESC_MAX_SIZE == 255, adding BUILD_BUG_ON(QUERY_DESC_SIZE != 255) and a comment may be sufficient.
Thanks,
Bart.
RE: [PATCH v5 4/4] ufs: core: Remove ufshcd_map_desc_id_to_length function
Quoted 5 lines
Should a check be added here that desc_buf[QUERY_DESC_LENGTH_OFFSET] <=
QUERY_DESC_MAX_SIZEto protect against firmware bugs? SinceQUERY_DESC_MAX_SIZE== 255, adding BUILD_BUG_ON(QUERY_DESC_SIZE !=
255)
and a comment may be sufficient.
Sorry - my bad.
In the previous review you mentioned that the patch looks good to you, hence the "Review by". Regarding your comment, we can do that, although I don't think we should cover for those FW basic bugs. Please let me know that you prefer.
Regards
Arthur
Quoted 41 lines
-----Original Message-----
From: Bart Van Assche <[email protected]>
Sent: Monday, December 12, 2022 2:19 AM
To: Arthur Simchaev <[email protected]>; [email protected]
Cc: [email protected]; [email protected]; linux- [email protected]
Subject: Re: [PATCH v5 4/4] ufs: core: Removeufshcd_map_desc_id_to_lengthfunctionCAUTION: This email originated from outside of Western Digital. Do not click on links or open attachments unless you recognize the sender and know that the content is safe.
Quoted 9 lines · Arthur Simchaev
There shouldn't be any restriction of the descriptor size (not the descriptor id for that matter) up to
QUERY_DESC_MAX_SIZE. According to the spec, the caller can use any descriptor size, and it is up to the device to return the actual size. Therefore there shouldn't be any sizes hardcoded in the kernel, nor any need to cache it, henceufshcd_map_desc_id_to_lengthfunction is redundant. Always read the descriptors withQUERY_DESC_MAX_SIZEsize.Reviewed-by Bart Van Assche <[email protected]>I have not yet replied with "Reviewed-by" to this patch so you are not yet allowed to add my Reviewed-by tag to this patch.
drivers/ufs/core/ufshcd.c int ufshcd_read_desc_param(struct ufs_hba *hba, · line 3419
Quoted 2 lines
drivers/ufs/core/ufshcd.c +2
-0+/* Update descriptor length */ + buff_len = desc_buf[QUERY_DESC_LENGTH_OFFSET ];Should a check be added here that desc_buf[QUERY_DESC_LENGTH_OFFSET] <=
QUERY_DESC_MAX_SIZEto protect against firmware bugs? SinceQUERY_DESC_MAX_SIZE== 255, adding BUILD_BUG_ON(QUERY_DESC_SIZE !=
255)
and a comment may be sufficient.Thanks,
Bart.
Bart Van Assche
reviewerRe: [PATCH v5 4/4] ufs: core: Remove ufshcd_map_desc_id_to_length function
Quoted 2 lines · Arthur Simchaev
Regarding your comment, we can do that, although I don't think we should cover for those FW basic bugs. Please let me know that you prefer.
I think this version is good enough. I will post my Reviewed-by.
Thanks,
Bart.
Re: [PATCH v5 4/4] ufs: core: Remove ufshcd_map_desc_id_to_length function
In reply to Arthur SimchaevQuoted 7 lines · Arthur Simchaev
There shouldn't be any restriction of the descriptor size (not the descriptor id for that matter) up to
QUERY_DESC_MAX_SIZE. According to the spec, the caller can use any descriptor size, and it is up to the device to return the actual size. Therefore there shouldn't be any sizes hardcoded in the kernel, nor any need to cache it, henceufshcd_map_desc_id_to_lengthfunction is redundant. Always read the descriptors withQUERY_DESC_MAX_SIZEsize.
Martin K. Petersen
AppliedArthur,
Quoted 4 lines
Read any descriptor with a maximum size of
QUERY_DESC_MAX_SIZE. According to the spec the device rerurns the actual size. Thus can improve code readability and save CPU cycles. While at it, cleanup few leftovers around the descriptor size parameter.
Applied to 6.3/scsi-staging, thanks!
Signature · 1 line
Martin K. Petersen Oracle Linux Engineering