I've got some patches to integrate...
Currently I have been working in two tracks.
1) Adding dirty page support to the page cache.
I had some preliminary discussions with Stephen Tweedie
2) Creating a proof of concept filesystem that uses dirty pages in the page cache, resides in swap, and is should be good for usage a Posix shared memory.
I believe I have found the last of my bugs, and it is now time to start integrating my work with the developmental kernel. I have heard that there is a feature freeze in place so my timing is horrible!
My patches are:
1. A patch to page_io.c that allows:
a) reading of swap pages directly into the page cache, with rw_swap_page_nocache
b) gives rw_swap_page_nocache a wait parameter. So it can do asynchronous operations.
c) uses brw_page directly for reading both swap partitions, and swap files, making life more uniform, allowing async operations on swap file etc.
2. A patch to swapfile.c.
The problem is that sys_swapoff assumes all swap pages are in mapped into processs. This is not necessarily true now with SYSV shared memory. And when my filesystem that resides in swap is mounted is very definentily not true.
The solution I have implemented is a some registration functions that maintain a linked list, so other parts of the kernel that use swap through rw_swap_page_nocache can have a chance to remove their pages from swap when swapoff is called.
3. A patch to filemap.c to that pages marked with a special dirty bit are written out in shrink_mmap.
This doesn't perform too well with the new memory allocator, that trys to keep lots of contigous memory. It pushes the swapping heavily case which is not yet handled very well.
On linux-2.0.32 it blasts writes through the page cache quite well.
Anyhow I thought I'd bounce off what I had off this list and see what people thought of my ideas. I started checking with Stepen Tweedie to see if he maintained the swap code and if I should send patches to him, or Linus. And he said it should bounce what I have off of this list, and to start talks about integrating it.
This is the end of the semester for me so I really won't have much time for at least two weeks :( My patches exist but need breaking up. Right now I just have one huge patch glob against 2.1.92 after I got everything working.
Eric
Hello Eric,
On 14 Apr 1998, Eric W. Biederman wrote:
...
Quoted 5 lines
Anyhow I thought I'd bounce off what I had off this list and see what people thought of my ideas. I started checking with Stepen Tweedie to see if he maintained the swap code and if I should send patches to him, or Linus. And he said it should bounce what I have off of this list, and to start talks about integrating it.
...
Hmmm, before I make any silly mumblings, would you mind [re]sending a copy of your patch to linux-mm? (I've misplaced the copy I had, and linuxhq's kernel archive isn't searchable right now.) Linus probably won't want to integrate them for 2.2, but perhaps we can start maintaining a set of patches reasonably 'in sync' until 2.3 when they could be integrated. Hopefully that way they we can get people to provide a reasonable amount of testing to the code.
Which reminds me: Stephen, what's the state of your irq and smp patches for page cache addition/removal. I'm getting a bit more free time now, so perhaps I can play with them a bit (maybe we should have a common cvs tree...).
-ben
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
"BLH" == Benjamin C R LaHaise <[email protected]> writes:
BLH> Hello Eric,
BLH> On 14 Apr 1998, Eric W. Biederman wrote:
BLH> ...
Quoted 5 lines
Quoted 5 lines
Anyhow I thought I'd bounce off what I had off this list and see what people thought of my ideas. I started checking with Stepen Tweedie to see if he maintained the swap code and if I should send patches to him, or Linus. And he said it should bounce what I have off of this list, and to start talks about integrating it.
BLH> ...
BLH> Hmmm, before I make any silly mumblings, would you mind [re]sending a copy BLH> of your patch to linux-mm?
Sending is right.
My filesytem code which also holds the patch is at:
http://www.npwt.net/~ebiederm/files/shmfs-0.1.008.tar.gz
I'm including a copy of my most recent patch against 2.1.92 at the end for completeness. Sorry it started out as a question who to....
Anyhow silly mumblings will be appreciated. :)
BLH> Linus probably won't want to
BLH> integrate them for 2.2, but perhaps we can start maintaining a set of BLH> patches reasonably 'in sync' until 2.3 when they could be integrated. BLH> Hopefully that way they we can get people to provide a reasonable amount BLH> of testing to the code.
begin 666 linux-2.1.92.x6.diff.gz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`:5[T;S MDR+3&E&1L$HU.I9AA@1TIA*16HHV=-R!)4U&>KO6B78)Y2=O_22D4KN^#&1$ MPU6+1^F-/G--.H].3!;L:.:9_8M(SXJ1,SO.!11XJUG/*Z83K8'WF=-;C8*> M7Z5E]7Q*>8VBA67(H:8KP5F"P?O<W^U1*'0]8>T$\I1'(9A_XE'F\,6CM%*I MS/HVIS+XT/$9$Q;BS^4OUM%9F_;?MC.-W;'=H]:=3"L+F.,R=NUB8$ZZ2'&D MRQHX?OA0:C8RJ+X*E'^G!J72>E.B-E#M!IC13WTZUE,^;02'M"^!-3'8/>]. MU>MUBM&2C.U6,3DSFYN-:JN570,.C8LPV'V42@;1L'I.:,CBJO!&NC"!Y?VN M2A:5EEJ(2QU/(B-'#FG5FD*&0,*N`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`EFDZJ7H9NN34DM-MEB5,B0?]7.!U6=XD&/!HSPP#3D)WTK= M:NND4:JEP(-[1>4<%C0##^P;NCFA[S"P",S6YDYU,Q*%)L^E2G3>GSN:J%3$ MV;LH:F/ER*/TRL:1?8S]LEHIL3'ELN5.?C(?N3:]5%S%$!DY,<B<>&HIWBPS MKDA&"P2CS4C?NC872T7$QYR"?1DT94J:@^)@GMK=S+@1<2^@'52N=>Q,'"++ ML\=A=<U)+=R5-($)NA=.3<C2_8K>,Y#SS?>"4$,FD*17TU3LZ;@ULJ\/XMU MXQ0-/#H?BO0?D^00<QP!?(R@8Q>]R(<;K^B#@[__AO3U-+I!'GTJ8I86"BZV MJ$>1Q*/.S\LFZ.-!5$=4UC&I(N;KX<CW9C<C/B#D_*(NH-<C)X![9SR&&]NW MW5A([email protected]+('AP?857_9QZ02,<9TTD,.["*[email protected]#-VWH#CAQ1Q MN%16@Z!.--8*U&7)7:+%FI,H2<Q,GO:\"GRQHXG5/]]7UB7N88GR<\3HQH M3XX=U19=!#/7UW>CW#"SWE7R'CH"Z?T8K6,-[(0V70A[R#O?$$;_;A_PRC M"3_]1,?)LVGH#?A:Q]\0-TK\K$2V]F+,=$O^P$JV282L-QP&BNQ6?E'?0? M**&7<7S98\8H]*0N9A7!>;+2ZO-S.9Y09F2[B^L@-(+RZ",LA=.Q3D#3J< M:F[OM=*7$';Y6A,]=G3LBB-"8'FWAG!WHK>SY.+<8W%$B[Q/9C<OK3[D@9I\
M.2\*>U>8KMA\A;<@PIG/@,/!X/5TI4K^NJL9?5FSACMR9(80M+U518C4IT ML5.J@V2AV;,!((3<F0EX_D#YE>_1CXE6B0DZEPL?^K]/N$78.8J8*ZCHOH7 MW1VTR56B,W>]6<!LGCK""V3!>J/>;-5W-^O01:<[4BY<WZ.K=92@7G%LQWK+ M[I/7!IM(Q33JJ'?UC3J.O%6']_[,]>"#[?3&&KD]D.,D"@WZ5$#B0CZRB-CK MT&K46T1M9WXV)A,<W]L/9_;]/3UJ2HR:HBOJN]-D#)_NJY\:&W5US6ECH^* M]MY1*.6)[>I[J(LNQA+_UIMT\6R]%6^8\6HP5QH$J"P2GH(H=:,YU!F]O?V MFWN2;X?PLG:!0<T:N5C$9P?SHQOR`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`E5%(U4I.@)@%5%8H6>]MLY+6M MK(V+4/][SVUFSNS..G:,D)``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`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`KG@8KL MY=?R0YB\"?IF.6V8E9P4SR+FMDY/?LCL9G!8QR:Z;FFTDP)8,X_QW8MS.V8I M5Y*'!:]&W")&&SW]!AL9$)F"#P%E]*"J8HPT1-@\Q6?8*>JM"J^&/V%FDVSD M&Y"GL9FYG&QF+B>*%SS:K$NV-.N2K<VZ9%NS+MG.K$M6FW4!IVLPZY+-S;JJ M\Z!*+H^A6>_L=32[TEA,'C86:\O4)%RE8!_F<E$EFA%_;W1.H*A\?S^=9XBL MOI\60<Y'AS63@'<AWQ).Q>R$DX)[*^1*==E4S*L9'U85^'47I5H#2VP+!2%9 M@C:UE2>A#UF(<LG2XDN>.D\[/JOQ)'U,*$)O>U@Q3@%!C]#(GZ!AZ?(WF MJ(:4EV5Z/RII&J"#P]^@9_VEDY"*1=P^'C[R'\DN\A_(P7?B&4N[GT)P>; MV^@)%B#-Q<C-9$8=TX.HX+R#4%1Z.H-%M5QERKZJ54L+3\_8=5L/26TCH5 M+%>(7*M4Q)VOV@^M/TP3:K'#G)^_V*26/Y72!4ZAMVIYW)/CE)5UE)2DROC M)/M*3+>XCH):]\,ZU(8L;;/@K4^,CH9O57,T@*O:6Q_A-;S+AF9_88O7_7[ M\,<,CIX?-85P]=B50=Q^MP^[email protected]=7D&7*%OEBYGF(Z6HSCE7RP%DIL1!'] MF9&4/TN2T,F_LW$^S.<^?ES/X==75D=Q$@EJ6PJ+YZ4X-C(ST0,/(QU?` MW^J]5\/EXHOV6O+Q(\B1$O@;J9%;&XF1FS=.)V@<6B'%O:,#C$3Q7Q*@E5PM M>.&@H@#?9VT]<UB/!:5TY/-=0X%L;)IDV)*".FPQYX;P-B8=H9!&OP##,FE> M8*U%.LI4T-D*79>,/,=RT)C80]=31V5V5.)RV/=&EJGS0ZI8SM&B*+Z&E3PC M(=DRMSHF-?R%,WQ24=P`/COC:-D^5SJCOXYT67%7="F"0:VLV[SAQ+;XKC#9 MKH7.:,R>>1IKK/YH$:?PTM^1!R8=C="9Q;F=_'6%LN((?.IPLZ"JH79_/_T' M\W$H3<^<8NG3T6*(P?MI0?8,O($Y&5F8;."R^=0EJ=?FLZ)M!SD4DF*4QJ5! MJ@2O<KB#I/'JHV3R+99]'L'$>C[09:0N_1L[N__Z,K1[@JDMU7?'5(1E.6 MJWYR:DFC\R._^$DQ&.=!*\;J4M4[3^V6*42>[K''A+A#;%75(-Y:[;:GGH M&*ALLA/?8)_5#O'&5FZKHS+_23>!NPLMPFJEGJUZCOCS=66;7UR2*V%92E M^*K_"W;K0"_54SFN=!/(3!.!Q'YW<_MYFKI6._GSW`[^':301K#J/^*P:_OS MQ-]P2A[/Y#NX&Y"Q\>N)7*QP%4)]I&6!9=U8PZ5B!C?"!MNZA(<I9]DP3\<4 MN'Q(C,S<-RC]Y>"(0.<#!Z5V#[QW<GYQ=G)V[.#1G\7A8KT?*S#CF+-R<G[U MQY_,%?H*.&[2+VD^1J.E$0&]R206!$VU[+-B6,S:.^@K[UV^?_O+R4[7=4R< MC=(;]+N#?C4-Y5JBDN+IHP27<H:9:LYX=\XJ?T2GYQ_>_KZI%B:[email protected] M&IAMW%C_<@//XW2P.>FC]C35X<*.GWXHD_52E[X;"CR9Q),Q7#\WF"91ZR0 M^\:T#_N84[W7P<+L%$UB=\8-G3F>9GN?K$=*J^)'26-[/1LOU,E-.[_!_1^ &4`<!:0``
`
end
I've got some patches to integrate, too...
In reply to Benjamin C.R. LaHaiseHi,
Quoted 4 lines · Benjamin C.R. LaHaise
Which reminds me: Stephen, what's the state of your irq and smp patches for page cache addition/removal. I'm getting a bit more free time now, so perhaps I can play with them a bit (maybe we should have a common cvs tree...).
Sorry, I've been offline, away in Colorado, for over a week, or I'd have answered this earlier...
The deferred pageout stuff (lazy page stealing) seems to work pretty well. It is rock solid when doing large compiles on a 6mb box, which stresses swapping pretty heavily. However, there is still a bug --- I can reproduce a copy-on-write violation occasionally when running one of my VM stress tests (I get a demand-zero page when I expect a non-zero value, so there's a pte getting lost in there somewhere). That's the only failure mode I can reproduce right now.
Right now I've got Linux Expo and Usenix deadlines approaching fast, so there's a limit to how much time I can spend debugging code which is not going to be integrated until 2.3. I can't promise to get the lazy page code debugged in the next couple of weeks. With the feature freeze in place, I guess there's a big question mark over whether this code can get included in 2.2. If it doesn't, then what we really need to concentrate on is the memory fragmentation issues in 2.1, not the new 2.2 features.
I spent a lot of time last week thinking about this, and there are a number of things we can do in 2.1 which will help enormously. I'll write them up in a day or two (right now I'm still catching up my mail backlog). We can definitely improve things without making major wholesale changes to the way VM works at the minute, even if it's not going to give us quite the performance or power that the really new code promises.
I guess that Linus really has to make the call whether or not we risk destabilising the existing VM by adding the new code now, or whether that is a 2.3 issue. If it is a 2.3 thing, then 2.2 has more urgent problems which need addressing as a higher priority than adding new functionality. We can still track the new VM code, but as a separate branch which won't be integrated until 2.3 development starts (much like the new net code which was integrated at the start of 1.3 development).
--Stephen