Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Fri, 6 Nov 2020 13:40:28 -0800
From:   Jaegeuk Kim <jaegeuk@kernel.org>
To:     Chao Yu <yuchao0@huawei.com>
Cc:     Eric Biggers <ebiggers@kernel.org>, linux-kernel@vger.kernel.org,
        linux-f2fs-devel@lists.sourceforge.net
Subject: Re: [f2fs-dev] [PATCH v3 2/2] f2fs: fix compat F2FS_IOC_{MOVE,
 GARBAGE_COLLECT}_RANGE
Message-ID: <20201106214028.GC1474936@google.com>
References: <20201105010934.16018-1-yuchao0@huawei.com>
 <20201106000550.GD2555324@gmail.com>
 <07454135-539d-a159-deb8-ff29df7e22de@huawei.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <07454135-539d-a159-deb8-ff29df7e22de@huawei.com>
Precedence: bulk

On 11/06, Chao Yu wrote:
> On 2020/11/6 8:05, Eric Biggers wrote:
> > This patch is marked 2/2, but it seems you sent it out on its own.  Patch series
> > are supposed to be resend in full; otherwise people can see just one patch and
> > have no context.
> 
> That's a historical problem, as in last many years, we (f2fs community) don't have
> other long-term reviewers except Jaegeuk and me, so we have unwritten rule: only
> resending changed patch in patchset.
> 
> IMO, that helps to skip traversing unchanged patches, and focusing reviewing on the
> real change lines, and certainly we have its context in mind.
> 
> Personally, I prefer to revise, resend or review patch{,es} of patchset have real
> change line in f2fs mailing list, anyway we can discuss about the rule here.

Chao, I think we need to change this to resend whole patch-set again, since
it's a bit difficult to catch which part of patches were the latest one.

> 
> > 
> > On Thu, Nov 05, 2020 at 09:09:34AM +0800, Chao Yu wrote:
> > > Eric reported a ioctl bug in below link:
> > > 
> > > https://lore.kernel.org/linux-f2fs-devel/20201103032234.GB2875@sol.localdomain/
> > > 
> > > That said, on some 32-bit architectures, u64 has only 32-bit alignment,
> > > notably i386 and x86_32, so that size of struct f2fs_gc_range compiled
> > > in x86_32 is 20 bytes, however the size in x86_64 is 24 bytes, binary
> > > compiled in x86_32 can not call F2FS_IOC_GARBAGE_COLLECT_RANGE successfully
> > > due to mismatched value of ioctl command in between binary and f2fs
> > > module, similarly, F2FS_IOC_MOVE_RANGE will fail too.
> > > 
> > > In this patch we introduce two ioctls for compatibility of above special
> > > 32-bit binary:
> > > - F2FS_IOC32_GARBAGE_COLLECT_RANGE
> > > - F2FS_IOC32_MOVE_RANGE
> > > 
> > 
> > It would be good to add a proper reported-by line, otherwise it's not clear who
> > "Eric" is (there are lots of Erics):
> > 
> > Reported-by: Eric Biggers <ebiggers@google.com>
> Sure, although I attached the link includes original report email, in where it
> points out who "Eric" is.
> 
> > 
> > > +static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range)
> > >   {
> > > -	struct inode *inode = file_inode(filp);
> > > -	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
> > > -	struct f2fs_gc_range range;
> > > +	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
> > >   	u64 end;
> > >   	int ret;
> > > +	if (unlikely(f2fs_cp_error(sbi)))
> > > +		return -EIO;
> > > +	if (!f2fs_is_checkpoint_ready(sbi))
> > > +		return -ENOSPC;
> > 
> > These two checkpoint-related checks weren't present in the original version.
> > Is that intentional?
> 
> Quoted
> 
> > It would be better to have __f2fs_ioc_gc_range() handle the f2fs_cp_error(),
> > f2fs_is_checkpoint_ready(), capable(), and f2fs_readonly() checks, so that they
> > don't have to be duplicated in the native and compat cases.
> 
> > Similarly for "move range".
> 
> I missed to check the detail, and just follow, I can clean up it.
> 
> > 
> > > +static int __f2fs_ioc_move_range(struct file *filp,
> > > +				struct f2fs_move_range *range)
> > >   {
> > > -	struct f2fs_move_range range;
> > > +	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
> > >   	struct fd dst;
> > >   	int err;
> > > +	if (unlikely(f2fs_cp_error(sbi)))
> > > +		return -EIO;
> > > +	if (!f2fs_is_checkpoint_ready(sbi))
> > > +		return -ENOSPC;
> > > +
> > 
> > Likewise here.
> > 
> > > diff --git a/include/uapi/linux/f2fs.h b/include/uapi/linux/f2fs.h
> > > index f00199a2e38b..8c14e88a9645 100644
> > > --- a/include/uapi/linux/f2fs.h
> > > +++ b/include/uapi/linux/f2fs.h
> > > @@ -5,6 +5,10 @@
> > >   #include <linux/types.h>
> > >   #include <linux/ioctl.h>
> > > +#ifdef __KERNEL__
> > > +#include <linux/compat.h>
> > > +#endif
> > > +
> > >   /*
> > >    * f2fs-specific ioctl commands
> > >    */
> > > @@ -65,6 +69,16 @@ struct f2fs_gc_range {
> > >   	__u64 len;
> > >   };
> > > +#if defined(__KERNEL__) && defined(CONFIG_COMPAT)
> > > +struct compat_f2fs_gc_range {
> > > +	u32 sync;
> > > +	compat_u64 start;
> > > +	compat_u64 len;
> > > +};
> > > +#define F2FS_IOC32_GARBAGE_COLLECT_RANGE	_IOW(F2FS_IOCTL_MAGIC, 11,\
> > > +						struct compat_f2fs_gc_range)
> > > +#endif
> > > +
> > >   struct f2fs_defragment {
> > >   	__u64 start;
> > >   	__u64 len;
> > > @@ -77,6 +91,17 @@ struct f2fs_move_range {
> > >   	__u64 len;		/* size to move */
> > >   };
> > > +#if defined(__KERNEL__) && defined(CONFIG_COMPAT)
> > > +struct compat_f2fs_move_range {
> > > +	u32 dst_fd;
> > > +	compat_u64 pos_in;
> > > +	compat_u64 pos_out;
> > > +	compat_u64 len;
> > > +};
> > > +#define F2FS_IOC32_MOVE_RANGE		_IOWR(F2FS_IOCTL_MAGIC, 9,	\
> > > +					struct compat_f2fs_move_range)
> > > +#endif
> > > +
> > >   struct f2fs_flush_device {
> > >   	__u32 dev_num;		/* device number to flush */
> > >   	__u32 segments;		/* # of segments to flush */
> > > -- 
> > 
> > Did you consider instead putting these compat definitions in an internal kernel
> > header, or even just in the .c file, to avoid cluttering up the UAPI header?
> 
> Better.
> 
> I can move them before their use.
> 
> > 
> > - Eric
> > .
> >