On Wed, Jun 9, 2010 at 2:58 PM, Boaz Harrosh <bharrosh@xxxxxxxxxxx> wrote:
> On 06/08/2010 07:19 AM, Fred Isaman wrote:
>> These will be used in the generic code. Set so they will compile away to
>> nothing if CONFIG_NFS_V4_1 not set.
>>
>> This requires kref_put to be under lock. See rule 3 of Documentation/kref.txt
>>
>
> I don't see "rule 3" in here. Please explain how?
3) If the code attempts to gain a reference to a kref-ed structure
without already holding a valid pointer, it must serialize access
where a kref_put() cannot occur during the kref_get(), and the
structure must remain valid during the kref_get().
This occurs every time we call pnfs_update_layout
>
> BTW: Even "rule 3" bad example with the lists, have a counter example in the Kernel
> with lists and searches that kref_put/get lockless. (By each element refing
> it's pear and taking the reference of the first one before search)
>
I don't follow this.
>> Signed-off-by: Fred Isaman <iisaman@xxxxxxxxxx>
>> ---
>> fs/nfs/pnfs.c | 45 ++++++++++++++++++++++++++++++++-------------
>> fs/nfs/pnfs.h | 44 +++++++++++++++++++++++++++++++++++++++++++-
>> 2 files changed, 75 insertions(+), 14 deletions(-)
>>
>> diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
>> index 836cb0f..a74a4b6 100644
>> --- a/fs/nfs/pnfs.c
>> +++ b/fs/nfs/pnfs.c
>> @@ -436,7 +436,25 @@ destroy_lseg(struct kref *kref)
>> PNFS_LD_IO_OPS(lseg->layout)->free_lseg(lseg);
>> }
>>
>> -static inline void
>> +static void
>> +put_lseg_locked(struct pnfs_layout_segment *lseg)
>> +{
>> + bool do_wake_up;
>> + struct nfs_inode *nfsi;
>> +
>> + if (!lseg)
>> + return;
>> +
>> + dprintk("%s: lseg %p ref %d valid %d\n", __func__, lseg,
>> + atomic_read(&lseg->kref.refcount), lseg->valid);
>> + do_wake_up = !lseg->valid;
>> + nfsi = PNFS_NFS_INODE(lseg->layout);
>> + kref_put(&lseg->kref, destroy_lseg);
>> + if (do_wake_up)
>> + wake_up(&nfsi->lo_waitq);
>> +}
>> +
>> +void
>> put_lseg(struct pnfs_layout_segment *lseg)
>> {
>> bool do_wake_up;
>> @@ -449,7 +467,9 @@ put_lseg(struct pnfs_layout_segment *lseg)
>> atomic_read(&lseg->kref.refcount), lseg->valid);
>> do_wake_up = !lseg->valid;
>> nfsi = PNFS_NFS_INODE(lseg->layout);
>> + lock_current_layout(nfsi);
>> kref_put(&lseg->kref, destroy_lseg);
>> + unlock_current_layout(nfsi);
>> if (do_wake_up)
>> wake_up(&nfsi->lo_waitq);
>> }
>> @@ -674,7 +694,7 @@ pnfs_free_layout(struct pnfs_layout_type *lo,
>> lseg, lseg->range.iomode, lseg->range.offset,
>> lseg->range.length);
>> list_del(&lseg->fi_list);
>> - put_lseg(lseg);
>> + put_lseg_locked(lseg);
>> }
>>
>> dprintk("%s:Return\n", __func__);
>> @@ -1033,7 +1053,7 @@ pnfs_has_layout(struct pnfs_layout_type *lo,
>> (lseg->valid || !only_valid)) {
>> ret = lseg;
>> if (take_ref)
>> - kref_get(&ret->kref);
>> + get_lseg(ret);
>> break;
>> }
>> if (cmp_layout(range, &lseg->range) > 0)
>> @@ -1053,7 +1073,7 @@ pnfs_has_layout(struct pnfs_layout_type *lo,
>> * returned to the caller.
>> */
>> int
>> -pnfs_update_layout(struct inode *ino,
>> +_pnfs_update_layout(struct inode *ino,
>> struct nfs_open_context *ctx,
>> u64 count,
>> loff_t pos,
>> @@ -1085,8 +1105,7 @@ pnfs_update_layout(struct inode *ino,
>> lseg = pnfs_has_layout(lo, &arg, take_ref, !take_ref);
>> if (lseg && !lseg->valid) {
>> if (take_ref)
>> - put_lseg(lseg);
>> -
>> + put_lseg_locked(lseg);
>> /* someone is cleaning the layout */
>> lseg = NULL;
>> result = -EAGAIN;
>> @@ -1262,7 +1281,7 @@ pnfs_layout_process(struct nfs4_pnfs_layoutget *lgp)
>> init_lseg(lo, lseg);
>> lseg->range = res->lseg;
>> if (lgp->lsegpp) {
>> - kref_get(&lseg->kref);
>> + get_lseg(lseg);
>> *lgp->lsegpp = lseg;
>> }
>>
>> @@ -1380,7 +1399,7 @@ pnfs_pageio_init_read(struct nfs_pageio_descriptor *pgio,
>> readahead_range(inode, pages, &loff, &count);
>>
>> if (count > 0) {
>> - status = pnfs_update_layout(inode, ctx, count,
>> + status = _pnfs_update_layout(inode, ctx, count,
>> loff, IOMODE_READ, NULL);
>> dprintk("%s virt update returned %d\n", __func__, status);
>> if (status != 0)
>> @@ -1438,7 +1457,7 @@ pnfs_update_layout_commit(struct inode *inode,
>> if (start == 0 && count == 0)
>> count = NFS4_MAX_UINT64;
>>
>> - status = pnfs_update_layout(inode, nfs_page->wb_context,
>> + status = _pnfs_update_layout(inode, nfs_page->wb_context,
>> count,
>> start,
>> IOMODE_RW,
>> @@ -1538,7 +1557,7 @@ pnfs_file_write(struct file *filp, const char __user *buf, size_t count,
>> goto out;
>>
>> /* Retrieve and set layout if not allready cached */
>> - status = pnfs_update_layout(inode,
>> + status = _pnfs_update_layout(inode,
>> context,
>> count,
>> *pos,
>> @@ -1580,7 +1599,7 @@ pnfs_writepages(struct nfs_write_data *wdata, int how)
>> args->offset);
>>
>> /* Retrieve and set layout if not allready cached */
>> - status = pnfs_update_layout(inode,
>> + status = _pnfs_update_layout(inode,
>> args->context,
>> args->count,
>> args->offset,
>> @@ -1681,7 +1700,7 @@ pnfs_readpages(struct nfs_read_data *rdata)
>> args->offset);
>>
>> /* Retrieve and set layout if not allready cached */
>> - status = pnfs_update_layout(inode,
>> + status = _pnfs_update_layout(inode,
>> args->context,
>> args->count,
>> args->offset,
>> @@ -1845,7 +1864,7 @@ pnfs_commit(struct nfs_write_data *data, int sync)
>> new one. If it was recalled we better commit the data first
>> before returning it, otherwise the data needs to be rewritten,
>> either with a new layout or to the MDS */
>> - result = pnfs_update_layout(data->inode,
>> + result = _pnfs_update_layout(data->inode,
>> NULL,
>> count,
>> first->wb_offset,
>> diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h
>> index 214d567..6326ed5 100644
>> --- a/fs/nfs/pnfs.h
>> +++ b/fs/nfs/pnfs.h
>> @@ -31,7 +31,8 @@ extern int pnfs4_proc_layoutreturn(struct nfs4_pnfs_layoutreturn *lrp, bool wait
>> /* pnfs.c */
>> extern const nfs4_stateid zero_stateid;
>>
>> -int pnfs_update_layout(struct inode *ino, struct nfs_open_context *ctx,
>> +void put_lseg(struct pnfs_layout_segment *lseg);
>> +int _pnfs_update_layout(struct inode *ino, struct nfs_open_context *ctx,
>> u64 count, loff_t pos, enum pnfs_iomode access_type,
>> struct pnfs_layout_segment **lsegpp);
>>
>> @@ -81,6 +82,12 @@ static inline int lo_fail_bit(u32 iomode)
>> NFS_INO_RW_LAYOUT_FAILED : NFS_INO_RO_LAYOUT_FAILED;
>> }
>>
>> +static inline void get_lseg(struct pnfs_layout_segment *lseg)
>> +{
>> + if (lseg)
>
> Really? That in my experience is a shoot in the foot.
>
> I don't believe any code that decided to get an lseg could get there without one.
> if so I want to crash.
>
> From all instances of get_lseg in this patch they already ask.
>
It is needed by one instance from the following patch. I can change this.
Fred
>> + kref_get(&lseg->kref);
>> +}
>> +
>> /* Return true if a layout driver is being used for this mountpoint */
>> static inline int pnfs_enabled_sb(struct nfs_server *nfss)
>> {
>> @@ -170,6 +177,23 @@ static inline int pnfs_return_layout(struct inode *ino,
>> return 0;
>> }
>>
>> +static inline int pnfs_update_layout(struct inode *ino,
>> + struct nfs_open_context *ctx,
>> + u64 count, loff_t pos, enum pnfs_iomode access_type,
>> + struct pnfs_layout_segment **lsegpp)
>> +{
>> + struct nfs_server *nfss = NFS_SERVER(ino);
>> +
>> + if (pnfs_enabled_sb(nfss))
>> + return _pnfs_update_layout(ino, ctx, count, pos,
>> + access_type, lsegpp);
>> + else {
>> + if (lsegpp)
>> + *lsegpp = NULL;
>> + return 0;
>> + }
>> +}
>> +
>> static inline int pnfs_get_write_status(struct nfs_write_data *data)
>> {
>> return data->pdata.pnfs_error;
>> @@ -190,6 +214,24 @@ static inline int pnfs_use_rpc(struct nfs_server *nfss)
>>
>> #else /* CONFIG_NFS_V4_1 */
>>
>> +static inline void get_lseg(struct pnfs_layout_segment *lseg)
>> +{
>> +}
>> +
>> +static inline void put_lseg(struct pnfs_layout_segment *lseg)
>> +{
>> +}
>> +
>> +static inline int
>> +pnfs_update_layout(struct inode *ino, struct nfs_open_context *ctx,
>> + u64 count, loff_t pos, enum pnfs_iomode access_type,
>> + struct pnfs_layout_segment **lsegpp)
>> +{
>> + if (lsegpp)
>> + *lsegpp = NULL;
>> + return 0;
>> +}
>> +
>> static inline enum pnfs_try_status
>> pnfs_try_to_read_data(struct nfs_read_data *data,
>> const struct rpc_call_ops *call_ops)
>
> Comments aside. Good needed stuff
> Boaz
> --
> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
