| Message ID | 20200602135409.GA59808@mwanda (mailing list archive) |
|---|---|
| State | Accepted |
| Delegated to: | Herbert Xu |
| Headers | show |
| Series | [v2] crypto: hisilicon - allow smaller reads in debugfs | expand |
Hi Dan, On 2020/6/2 21:54, Dan Carpenter wrote: > Originally this code rejected any read less than 256 bytes. There > is no need for this artificial limit. We should just use the normal > helper functions to read a string from the kernel. > > Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> > --- > v2: Use simple_read_from_buffer(). The v1 was slightly half arsed > because I left the original check for: > > if (*pos) > return 0; > > So it could result in partial reads. The new code means that if you > want to read the buffer one byte at a time, that's fine or if you want > to read it in one 256 byte chunk that's also fine. Plus it deletes 21 > lines of code and is a lot cleaner. > In fact, In our original design, we do not hope the user do the partial reads. Thank you for your work, but I still insist on adding this limit. Thanks, Shukun > drivers/crypto/hisilicon/qm.c | 33 ++++++--------------------------- > 1 file changed, 6 insertions(+), 27 deletions(-) > > diff --git a/drivers/crypto/hisilicon/qm.c b/drivers/crypto/hisilicon/qm.c > index 9bb263cec6c30..13ccb9e29a2e1 100644 > --- a/drivers/crypto/hisilicon/qm.c > +++ b/drivers/crypto/hisilicon/qm.c > @@ -1064,19 +1064,10 @@ static ssize_t qm_cmd_read(struct file *filp, char __user *buffer, > char buf[QM_DBG_READ_LEN]; > int len; > > - if (*pos) > - return 0; > - > - if (count < QM_DBG_READ_LEN) > - return -ENOSPC; > + len = scnprintf(buf, QM_DBG_READ_LEN, "%s\n", > + "Please echo help to cmd to get help information"); > > - len = snprintf(buf, QM_DBG_READ_LEN, "%s\n", > - "Please echo help to cmd to get help information"); > - > - if (copy_to_user(buffer, buf, len)) > - return -EFAULT; > - > - return (*pos = len); > + return simple_read_from_buffer(buffer, count, pos, buf, len); > } > > static void *qm_ctx_alloc(struct hisi_qm *qm, size_t ctx_size, > @@ -2691,24 +2682,12 @@ static ssize_t qm_status_read(struct file *filp, char __user *buffer, > { > struct hisi_qm *qm = filp->private_data; > char buf[QM_DBG_READ_LEN]; > - int val, cp_len, len; > - > - if (*pos) > - return 0; > - > - if (count < QM_DBG_READ_LEN) > - return -ENOSPC; > + int val, len; > > val = atomic_read(&qm->status.flags); > - len = snprintf(buf, QM_DBG_READ_LEN, "%s\n", qm_s[val]); > - if (!len) > - return -EFAULT; > - > - cp_len = copy_to_user(buffer, buf, len); > - if (cp_len) > - return -EFAULT; > + len = scnprintf(buf, QM_DBG_READ_LEN, "%s\n", qm_s[val]); > > - return (*pos = len); > + return simple_read_from_buffer(buffer, count, pos, buf, len); > } > > static const struct file_operations qm_status_fops = { >
On Fri, Jun 05, 2020 at 09:19:53AM +0800, Shukun Tan wrote: > Hi Dan, > > On 2020/6/2 21:54, Dan Carpenter wrote: > > Originally this code rejected any read less than 256 bytes. There > > is no need for this artificial limit. We should just use the normal > > helper functions to read a string from the kernel. > > > > Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> > > --- > > v2: Use simple_read_from_buffer(). The v1 was slightly half arsed > > because I left the original check for: > > > > if (*pos) > > return 0; > > > > So it could result in partial reads. The new code means that if you > > want to read the buffer one byte at a time, that's fine or if you want > > to read it in one 256 byte chunk that's also fine. Plus it deletes 21 > > lines of code and is a lot cleaner. > > > > In fact, In our original design, we do not hope the user do the partial reads. > Thank you for your work, but I still insist on adding this limit. This not how POSIX filesystems work... :( Last time you said that this literally breaks cat. This doesn't break anything if the user chooses not to read a single byte at a time. That's obviously a crazy way to read a file. It just allows them to if they want. Or if they want to read 256 bytes at a time then that also works. My patch makes *everything* work. regards, dan carpenter
On Tue, Jun 02, 2020 at 04:54:09PM +0300, Dan Carpenter wrote: > Originally this code rejected any read less than 256 bytes. There > is no need for this artificial limit. We should just use the normal > helper functions to read a string from the kernel. > > Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> > --- > v2: Use simple_read_from_buffer(). The v1 was slightly half arsed > because I left the original check for: > > if (*pos) > return 0; > > So it could result in partial reads. The new code means that if you > want to read the buffer one byte at a time, that's fine or if you want > to read it in one 256 byte chunk that's also fine. Plus it deletes 21 > lines of code and is a lot cleaner. > > drivers/crypto/hisilicon/qm.c | 33 ++++++--------------------------- > 1 file changed, 6 insertions(+), 27 deletions(-) Patch applied. Thanks.
diff --git a/drivers/crypto/hisilicon/qm.c b/drivers/crypto/hisilicon/qm.c index 9bb263cec6c30..13ccb9e29a2e1 100644 --- a/drivers/crypto/hisilicon/qm.c +++ b/drivers/crypto/hisilicon/qm.c @@ -1064,19 +1064,10 @@ static ssize_t qm_cmd_read(struct file *filp, char __user *buffer, char buf[QM_DBG_READ_LEN]; int len; - if (*pos) - return 0; - - if (count < QM_DBG_READ_LEN) - return -ENOSPC; + len = scnprintf(buf, QM_DBG_READ_LEN, "%s\n", + "Please echo help to cmd to get help information"); - len = snprintf(buf, QM_DBG_READ_LEN, "%s\n", - "Please echo help to cmd to get help information"); - - if (copy_to_user(buffer, buf, len)) - return -EFAULT; - - return (*pos = len); + return simple_read_from_buffer(buffer, count, pos, buf, len); } static void *qm_ctx_alloc(struct hisi_qm *qm, size_t ctx_size, @@ -2691,24 +2682,12 @@ static ssize_t qm_status_read(struct file *filp, char __user *buffer, { struct hisi_qm *qm = filp->private_data; char buf[QM_DBG_READ_LEN]; - int val, cp_len, len; - - if (*pos) - return 0; - - if (count < QM_DBG_READ_LEN) - return -ENOSPC; + int val, len; val = atomic_read(&qm->status.flags); - len = snprintf(buf, QM_DBG_READ_LEN, "%s\n", qm_s[val]); - if (!len) - return -EFAULT; - - cp_len = copy_to_user(buffer, buf, len); - if (cp_len) - return -EFAULT; + len = scnprintf(buf, QM_DBG_READ_LEN, "%s\n", qm_s[val]); - return (*pos = len); + return simple_read_from_buffer(buffer, count, pos, buf, len); } static const struct file_operations qm_status_fops = {
Originally this code rejected any read less than 256 bytes. There is no need for this artificial limit. We should just use the normal helper functions to read a string from the kernel. Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> --- v2: Use simple_read_from_buffer(). The v1 was slightly half arsed because I left the original check for: if (*pos) return 0; So it could result in partial reads. The new code means that if you want to read the buffer one byte at a time, that's fine or if you want to read it in one 256 byte chunk that's also fine. Plus it deletes 21 lines of code and is a lot cleaner. drivers/crypto/hisilicon/qm.c | 33 ++++++--------------------------- 1 file changed, 6 insertions(+), 27 deletions(-)