Re: [Nbd] [Qemu-devel] [PATCH v5 13/14] nbd: Implement NBD_CMD_WRITE_ZEROES on server
- To: Paolo Bonzini <pbonzini@...696...>
- Cc: Kevin Wolf <kwolf@...696...>, "nbd-general@lists.sourceforge.net" <nbd-general@lists.sourceforge.net>, qemu-block@...530..., qemu-devel@...530..., Max Reitz <mreitz@...696...>
- Subject: Re: [Nbd] [Qemu-devel] [PATCH v5 13/14] nbd: Implement NBD_CMD_WRITE_ZEROES on server
- From: Fam Zheng <famz@...696...>
- Date: Wed, 20 Jul 2016 15:38:36 +0800
- Message-id: <20160720073836.GF10539@...2425...>
- In-reply-to: <913397c9-6edc-2561-3d2e-e32032f9db22@...696...>
- References: <1468901281-22858-1-git-send-email-eblake@...696...> <1468901281-22858-14-git-send-email-eblake@...696...> <20160719062131.GG18103@...2425...> <578E4708.5080308@...696...> <aad94967-fd7b-4377-ad40-3f07d67b5c32@...696...> <20160720033402.GA7641@...2425...> <578EF446.70202@...696...> <20160720043709.GA10539@...2425...> <913397c9-6edc-2561-3d2e-e32032f9db22@...696...>
On Wed, 07/20 09:09, Paolo Bonzini wrote:
>
>
> On 20/07/2016 06:37, Fam Zheng wrote:
> > Yes, you are right about this, I was confused because "qemu-img map" does not
> > report this allocation state after zero write. (No idea why SEEK_DATA doesn't
> > hit the fallocate'ed area.)
>
> Apparently it's because it's zeroed.
>
> $ fallocate -z -o 10485760 -l 10485760 test
> $ fallocate -p -o 49152000 -l 10485760 test
> $ fallocate -o 49152000 -l 10485760 test
> $ fallocate -p -o 65536000 -l 10485760 test
>
> Now we have:
>
> - a zero area at 10240K..20480K
>
> - an hole+allocated area at 48000K..59240K
>
> - a hole at 64000K..74240K
>
> $ qemu-img map test
> Offset Length Mapped to File
> 0 0xa00000 0 test << ends at 10240K
> 0x1400000 0x1ae0000 0x1400000 test << ends at 48000K
> 0x38e0000 0x5a0000 0x38e0000 test << ends at 64000K
> 0x4880000 0x1b80000 0x4880000 test
Hah? I'm apparently missing something. I can't see these entries:
> fam@...2426...:/var/tmp$ cat /tmp/sh
> touch test
> fallocate -z -o 10485760 -l 10485760 test
> fallocate -p -o 49152000 -l 10485760 test
> fallocate -o 49152000 -l 10485760 test
> fallocate -p -o 65536000 -l 10485760 test
> qemu-img map test
> fam@...2426...:/var/tmp$ sh /tmp/sh
> Offset Length Mapped to File
(I'm using Fedora 24 but I've also tried RHEL 7.)
Fam
>
> So "qemu-img map" hides both zeroed and hole areas. With the JSON format
> we get more information:
>
> $ qemu-img map --output=json test
> [{ "start": 0, "length": 10485760, "depth": 0, "zero": false, "data": true, "offset": 0},
> { "start": 10485760, "length": 10485760, "depth": 0, "zero": true, "data": false, "offset": 10485760},
> { "start": 20971520, "length": 28180480, "depth": 0, "zero": false, "data": true, "offset": 20971520},
> { "start": 49152000, "length": 10485760, "depth": 0, "zero": true, "data": false, "offset": 49152000},
> { "start": 59637760, "length": 5898240, "depth": 0, "zero": false, "data": true, "offset": 59637760},
> { "start": 65536000, "length": 10485760, "depth": 0, "zero": true, "data": false, "offset": 65536000},
> { "start": 76021760, "length": 28835840, "depth": 0, "zero": false, "data": true, "offset": 76021760}]
>
> Both zeroed and holes are reported as "zero": true, "data": false. This
> limitation stems from the fact that we cannot use FIEMAP.
>
> Paolo
Reply to: