Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

[.data.table crashes with segfault while grouping with more than 79 threads #5077

Closed
bennet0496 opened this issue Jul 21, 2021 · 1 comment · Fixed by #5087
Closed

[.data.table crashes with segfault while grouping with more than 79 threads #5077

bennet0496 opened this issue Jul 21, 2021 · 1 comment · Fixed by #5087
Labels
bug High
Milestone
1.15.0

Comments

@bennet0496
Copy link
Contributor

I'm working at the IT department in a research institute, and recently one of our scientists came to me with mysteriously crashing R sessions on one (and only one) of our machines (designated for use with R and Rstudio) at the institute. While investigating the problem, I found that the main and deciding difference between the machines he tried it on is the CPU and more specifically the number of cores of the CPUs. The machine the code was working on was a 4 socket, 16 core machine and the machine where the crashes occurred was a 2 socket, 265 core (with SMT) machine. Further investigation revealed, this was important, because when importing data.table, by default 50% of the available cores are chosen for multithreaded operations.

From playing around with setDTthreads(threads = N), I found that N = 79 seamed to be the magical threshold for the crashes to occur. Below this number the code consistently works, above this number the code will crash in most (but not all) cases.

The error first occured with R 4.0.2 and data.table 1.13.0. But because #4892 looked somewhat similar, I also tested it with the latest development version 1.14.1.

Please find blow the crashing code, from what he told me, the code is trying to archive the follwing:

What we are calculating is what is called “the nearest neighbour distance” (it is a term). The “AgentID” is an ID given to each “person” to identify the data belong to them. For each “Time” and for each “AgentID”, we are calculating what is the distance of closest person.

Minimal reproducible example

library(data.table)

M  <- 160
M2 <- M / 2
M4 <- M / 4

J  <- 10001

setDTthreads(threads = 96)

d2 <- data.table(
        Time=rep(1:J, each=M),
        AgentID=rep(1:M4, times=4*J*M),
        i_phase=sample(0:1, J*M, replace=T)*2-1,
        State=sample(0:99,J*M, replace=T),
        SimulationID=sample(1:4, J*M, replace=T),
        N=rep(M4, times=J*M),
        discountFactor=rep(0, times=J*M)
)

findNNDistance <- function(x){
    return(Rfast::nth(x,2,descending = F))
}

d2[, nnDistance_inphase := apply(as.matrix(dist(State, method = "euclidean")), 1, FUN=findNNDistance), by=.(Time,i_phase,N,SimulationID,discountFactor), verbose=T]

error

The spcifics of the segfault vary, depending on the number of threads, so here are the error for a variaty of thread numbers

the result for `N = 80` threads 
Argument 'by' after substitute: .(Time, i_phase, N, SimulationID, discountFactor)
Detected that j uses these columns: [State]
Finding groups using forderv ... forder.c received 256025600 rows and 5 columns
12.3s elapsed (00:03:20 cpu) 
Finding group sizes from the positions (can be avoided to save RAM) ... 0.002s elapsed (0.001s cpu) 
Getting back original order ... forder.c received a vector type 'integer' length 80008
*** Error in `/usr/local/R/4.0.2/lib64/R/bin/exec/R': double free or corruption (!prev): 0x00000000040ae990 ***
Aborted (core dumped)
the result for `N = 96` threads 
Argument 'by' after substitute: .(Time, i_phase, N, SimulationID, discountFactor)
Detected that j uses these columns: [State]
Finding groups using forderv ... forder.c received 256025600 rows and 5 columns
13.2s elapsed (00:04:12 cpu) 
Finding group sizes from the positions (can be avoided to save RAM) ... 0.001s elapsed (0.001s cpu) 
Getting back original order ... forder.c received a vector type 'integer' length 80008

 *** caught segfault ***

 *** caught segfault ***
address 0x64ff818, cause 'memory not mapped'

 *** caught segfault ***
address 0x623f844, cause 'memory not mapped'

 *** caught segfault ***
address 0x61bf84c, cause 'memory not mapped'
address 0x64bf81c, cause 'memory not mapped'

 *** caught segfault ***
address 0x62ff838, cause 'memory not mapped'

 *** caught segfault ***
address 0x61ff848, cause 'memory not mapped'

 *** caught segfault ***
address 0x63ff828, cause 'memory not mapped'

 *** caught segfault ***
address 0x647f820, cause 'memory not mapped'

 *** caught segfault ***
address 0x627f840, cause 'memory not mapped'

 *** caught segfault ***
address 0x643f824, cause 'memory not mapped'

 *** caught segfault ***
address 0x62bf83c, cause 'memory not mapped'

 *** caught segfault ***
address 0x63bf82c, cause 'memory not mapped'

 *** caught segfault ***
address 0x633f834, cause 'memory not mapped'

 *** caught segfault ***
address 0x653f814, cause 'memory not mapped'

 *** caught segfault ***
address 0x637f830, cause 'memory not mapped'
Error: no more error handlers available (recursive errors?); invoking 'abort' restart
Error: no more error handlers available (recursive errors?); invoking 'abort' restart
Error: bad value
Error: no more error handlers available (recursive errors?); invoking 'abort' restart
Error: no more error handlers available (recursive errors?); invoking 'abort' restart
Error: no more error handlers available (recursive errors?); invoking 'abort' restart
Error: no more error handlers available (recursive errors?); invoking 'abort' restart
Error: no more error handlers available (recursive errors?); invoking 'abort' restart
Error: no more error handlers available (recursive errors?); invoking 'abort' restart
Error: bad value
Error: no more error handlers available (recursive errors?); invoking 'abort' restart
*** stack smashing detected ***: <unknown> terminated
> Segmentation fault (core dumped)
the result for `N = 128` threads 
Argument 'by' after substitute: .(Time, i_phase, N, SimulationID, discountFactor)
Detected that j uses these columns: [State]
Finding groups using forderv ... forder.c received 256025600 rows and 5 columns
35.9s elapsed (00:49:31 cpu) 
Finding group sizes from the positions (can be avoided to save RAM) ... 0.001s elapsed (0.000s cpu) 
Getting back original order ... forder.c received a vector type 'integer' length 80008

*** caught segfault ***
address 0x7f7178721000, cause 'invalid permissions'

Traceback:
1: forderv(firstofeachgroup)
2: `[.data.table`(d2, , `:=`(nnDistance_inphase, apply(as.matrix(dist(State,     method = "euclidean")), 1, FUN = findNNDistance)), by = .(Time,     i_phase, N, SimulationID, discountFactor), verbose = T)
3: d2[, `:=`(nnDistance_inphase, apply(as.matrix(dist(State, method = "euclidean")),     1, FUN = findNNDistance)), by = .(Time, i_phase, N, SimulationID,     discountFactor), verbose = T]

Possible actions:
1: abort (with core dump, if enabled)
2: normal R exit
3: exit R without saving workspace
4: exit R saving workspace
Selection: 1
R is aborting now ...
Segmentation fault (core dumped)

Because the M = 160 seamed related to the threshold of 79/80 cores, I also tried other M, specifically M = 100 but this error threshold still remains the same.

Output of sessionInfo()

R version 4.0.2 (2020-06-22)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: openSUSE Leap 15.2

Matrix products: default
BLAS:   /usr/local/R/4.0.2/lib64/R/lib/libRblas.so
LAPACK: /usr/local/R/4.0.2/lib64/R/lib/libRlapack.so

locale:
 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
 [3] LC_TIME=de_DE.UTF-8        LC_COLLATE=en_US.UTF-8    
 [5] LC_MONETARY=de_DE.UTF-8    LC_MESSAGES=en_US.UTF-8   
 [7] LC_PAPER=de_DE.UTF-8       LC_NAME=C                 
 [9] LC_ADDRESS=C               LC_TELEPHONE=C            
[11] LC_MEASUREMENT=de_DE.UTF-8 LC_IDENTIFICATION=C       

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] data.table_1.14.1

loaded via a namespace (and not attached):
[1] compiler_4.0.2
@bennet0496
Copy link
Contributor Author

bennet0496 commented Jul 23, 2021 •
edited
Loading

I did some more testing and it turns out the double free for N = 80 is caused by the cleanup function in forder.c:88, while freeing the global TMP pointer. I wasn"t able to figure out where the first call to a free would have been or if TMP even had a value in the first place, but I guess sourrounding it with a NULL check would do the trick there.

GDB Backtrace for `N = 80` 
#0  0x00007ffff6d2f420 in raise () from /lib64/libc.so.6
No symbol table info available.
#1  0x00007ffff6d30a01 in abort () from /lib64/libc.so.6
No symbol table info available.
#2  0x00007ffff6d72877 in __libc_message () from /lib64/libc.so.6
No symbol table info available.
#3  0x00007ffff6d79093 in malloc_printerr () from /lib64/libc.so.6
No symbol table info available.
#4  0x00007ffff6d7a999 in _int_free () from /lib64/libc.so.6
No symbol table info available.
#5  0x00007ffff05eecde in cleanup () at forder.c:88
No locals.
#6  0x00007ffff05f2541 in forder (DT=0x30c4190, by=0x30c41c8, retGrpArg=<optimized out>, sortGroupsArg=<optimized out>, ascArg=0x30d3710, naArg=<optimized out>) at forder.c:771
        n_protect = 3
        verbose = true
        n_cplx = <optimized out>
        ans = 0x3576180
        ncol = 1
        keyAlloc = <optimized out>
        spare = <optimized out>
        Rf_isReal = <optimized out>
        complexRerun = <optimized out>
        CplxPart = 0x6215a0
#7  0x00007ffff783bb64 in R_doDotCall (ofun=<optimized out>, nargs=<optimized out>, cargs=<optimized out>, call=0x2ac6128) at dotcode.c:614
        fun = <optimized out>
        retval = <optimized out>
#8  0x00007ffff787b38a in bcEval (body=body@entry=0x2ad4de0, rho=rho@entry=0x2a7e350, useCache=useCache@entry=TRUE) at eval.c:7655
        cargs = {0x3527f20, 0x6215a0, 0x30d36a0, 0x30d36d8, 0x30d3710, 0x30d3748, 0x80, 0x7ffff6d7def7 <malloc+199>, 0x0, 0xffffffff, 0x7fffffff9ea0, 0x80, 0x13, 0x7ffff78c4ec4 <Rf_allocVector3+1796>, 
          0x7fffffff9ea0, 0x7fff00000007}
        vmax = 0x0
        val = <optimized out>
        call = 0x2ac6128
        nargs = 6
        ofun = 0x7ffff05f1b90 <forder>
        sym = <optimized out>
        args = <optimized out>
        op = <optimized out>
        retvalue = <optimized out>
        constants = 0xccca50
        pc = <optimized out>
        codebase = 0x294e1e0
        oldntop = <optimized out>
        evalcount = 391
        oldsrcref = <optimized out>
        oldbcintactive = <optimized out>
        oldbcbody = <optimized out>
        oldbcpc = <optimized out>
        currentpc = 0x294e5d0
        which = 0
        init = <optimized out>
        old_byte_code = <optimized out>
        ibcl_oldptop = 0x7ffff3b10020
        vcache = 0x7ffff3b10090
        smallcache = <optimized out>
        loop = <optimized out>
#9  0x00007ffff78855f0 in Rf_eval (e=0x2ad4de0, rho=rho@entry=0x2a7e350) at eval.c:723
        op = <optimized out>
        tmp = <optimized out>
        evalcount = 508
        bcintactivesave = 1
        srcrefsave = 0x6215a0
        depthsave = 2
#10 0x00007ffff788736f in R_execClosure (call=call@entry=0x2e8cc68, newrho=newrho@entry=0x2a7e350, sysparent=<optimized out>, rho=rho@entry=0x2d18570, arglist=arglist@entry=0x2a7e1c8, op=op@entry=0x2ad48d8) at eval.c:1888
        body = 0x2ad4de0
        cntxt = {nextcontext = 0x7fffffffb710, callflag = 12, cjmpbuf = {{__jmpbuf = {44912096, -69324313296529690, 140737351868760, 44910808, 140737351868704, 6428064, 69330512277772006, 69347660491101926}, __mask_was_saved = 0, __saved_mask = {__val = {7, 140737351869784, 1701147977, 19819992, 80008, 256025600, 140724853501968, 53199584, 140737351868760, 140724853501968, 6425936, 53199584, 140737351869872, 55738144, 140737345629784, 140737351868760}}}}, cstacktop = 16, evaldepth = 2, promargs = 0x2a7e1c8, callfun = 0x2ad48d8, sysparent = 0x2d18570, call = 0x2e8cc68, cloenv = 0x2a7e350, conexit = 0x6215a0, cend = 0x0, cenddata = 0x7ffff78bab14 <Rf_matchArgs_NR+1364>, vmax = 0x0, intsusp = 0, gcenabled = 1, bcintactive = 1, bcbody = 0x2f8ee50, bcpc = 0x7fffffffaf78, handlerstack = 0x6215a0, restartstack = 0x6215a0, prstack = 0x0, nodestack = 0x7ffff3b10080, bcprottop = 0x7ffff3b10020, srcref = 0x621488, browserfinish = 0, returnValue = 0x0, jumptarget = 0x0, jumpmask = 0}
        dbg = FALSE
#11 0x00007ffff7888137 in Rf_applyClosure (call=call@entry=0x2e8cc68, op=op@entry=0x2ad48d8, arglist=arglist@entry=0x2a7e1c8, rho=rho@entry=0x2d18570, suppliedvars=<optimized out>) at eval.c:1814
        formals = <optimized out>
        actuals = <optimized out>
        savedrho = <optimized out>
        newrho = 0x2a7e350
        f = 0x6215a0
        a = 0x6215a0
        is_getter_call = FALSE
        val = <optimized out>
#12 0x00007ffff787c519 in bcEval (body=body@entry=0x2f8ee50, rho=rho@entry=0x2d18570, useCache=useCache@entry=TRUE) at eval.c:7067
        fun = 0x2ad48d8
        call = 0x2e8cc68
        value = 0x0
        flag = <optimized out>
        args = <optimized out>
        retvalue = <optimized out>
        constants = 0x2631f50
        pc = 0x2c58438
        codebase = 0x2c403e0
        oldntop = <optimized out>
        evalcount = 391
        oldsrcref = <optimized out>
        oldbcintactive = <optimized out>
        oldbcbody = <optimized out>
        oldbcpc = <optimized out>
        currentpc = 0x2c58420
        which = 0
        init = <optimized out>
        old_byte_code = <optimized out>
        ibcl_oldptop = 0x7ffff3b0f010
        vcache = 0x7ffff3b0f020
        smallcache = <optimized out>
        loop = <optimized out>
#13 0x00007ffff78855f0 in Rf_eval (e=0x2f8ee50, rho=rho@entry=0x2d18570) at eval.c:723
        op = <optimized out>
        tmp = <optimized out>
        evalcount = 508
        bcintactivesave = 0
        srcrefsave = 0x6215a0
        depthsave = 1
#14 0x00007ffff788736f in R_execClosure (call=call@entry=0x2d18148, newrho=newrho@entry=0x2d18570, sysparent=<optimized out>, rho=rho@entry=0x2f8d330, arglist=arglist@entry=0x2f92f78, op=op@entry=0x2f8d8a8) at eval.c:1888
        body = 0x2f8ee50
        cntxt = {nextcontext = 0x7fffffffbac0, callflag = 12, cjmpbuf = {{__jmpbuf = {49868368, -69324529462569242, 140737351868760, 49862824, 140737351868704, 6428064, 69330512674133734, 69347660491101926}, __mask_was_saved = 0, __saved_mask = {__val = {4421, 0, 0, 140737351869784, 140737346569376, 140737351869784, 140737347655273, 0, 140737351869176, 9627792, 2, 0, 140737346572518, 8371735909806386779, 139639678329441, 6428064}}}}, cstacktop = 15, evaldepth = 1, promargs = 0x2f92f78, callfun = 0x2f8d8a8, sysparent = 0x6595c8, call = 0x2d18148, cloenv = 0x2d18570, conexit = 0x2d11638, cend = 0x0, cenddata = 0x0, vmax = 0x0, intsusp = 0, gcenabled = 1, bcintactive = 0, bcbody = 0x0, bcpc = 0x0, handlerstack = 0x6215a0, restartstack = 0x6215a0, prstack = 0x0, nodestack = 0x7ffff3b0f010, bcprottop = 0x7ffff3b0f010, srcref = 0x6215a0, browserfinish = 0, returnValue = 0x0, jumptarget = 0x0, jumpmask = 0}
        dbg = FALSE
#15 0x00007ffff7888137 in Rf_applyClosure (call=<optimized out>, op=<optimized out>, arglist=<optimized out>, rho=0x2f8d330, suppliedvars=<optimized out>) at eval.c:1814
        formals = <optimized out>
        actuals = <optimized out>
        savedrho = <optimized out>
        newrho = 0x2d18570
        f = 0x6215a0
        a = 0x6215a0
        is_getter_call = FALSE
        val = <optimized out>
#16 0x00007ffff78c9463 in applyMethod (call=call@entry=0x2d18148, op=op@entry=0x2f8d8a8, args=<optimized out>, rho=rho@entry=0x2f8d330, newvars=newvars@entry=0x2d22398) at objects.c:118
        ans = <optimized out>
#17 0x00007ffff78ca203 in dispatchMethod (sxp=sxp@entry=0x2f8d8a8, dotClass=dotClass@entry=0x30acb88, cptr=cptr@entry=0x7fffffffbac0, method=method@entry=0x26c2e88, generic=0x7ffff79d1669 "[", rho=0x2f8d330, callrho=0x6595c8, defrho=0x621530, op=<optimized out>, op=<optimized out>) at objects.c:436
        newvars = 0x2d22398
        newcall = <optimized out>
        matchedarg = <optimized out>
        ans = <optimized out>
#18 0x00007ffff78ca6b4 in Rf_usemethod (generic=generic@entry=0x7ffff79d1669 "[", obj=obj@entry=0x92e890, call=call@entry=0x2f92df0, args=args@entry=0x2f92f78, rho=rho@entry=0x2f8d330, callrho=callrho@entry=0x6595c8, defrho=0x621530, ans=0x7fffffffbcc8) at objects.c:476
        vmax = <optimized out>
        ss = <optimized out>
        klass = 0x30acb88
        method = 0x26c2e88
        sxp = 0x2f8d8a8
        op = 0x62a5e0
        i = 0
        nclass = <optimized out>
        cptr = 0x7fffffffbac0
#19 0x00007ffff788c651 in Rf_DispatchOrEval (call=call@entry=0x2f92df0, op=op@entry=0x62a5e0, generic=generic@entry=0x7ffff79d1669 "[", args=0x2f92ed0, rho=rho@entry=0x6595c8, ans=ans@entry=0x7fffffffbcc8, dropmissing=0, argsevald=0) at eval.c:3610
        cntxt = {nextcontext = 0x7ffff7dd6360 <R_Toplevel>, callflag = 20, cjmpbuf = {{__jmpbuf = {140733193388032, 140737351869176, 140737351869784, -69330188288551194, 140737351868760, 38117464, 140737351868704, 6428064}, __mask_was_saved = -41314586, __saved_mask = {__val = {69347660491101926, 0, 0, 0, 6657480, 140737488338816, 140737351934547, 1, 0, 0, 140737345000376, 140737488338720, 140737351963754, 0, 0, 1}}}}, cstacktop = 7, evaldepth = 1, promargs = 0x2f92f78, callfun = 0x62a5e0, sysparent = 0x6595c8, call = 0x2f92df0, cloenv = 0x2f8d330, conexit = 0x6215a0, cend = 0x0, cenddata = 0x6215a0, vmax = 0x0, intsusp = 0, gcenabled = 1, bcintactive = 0, bcbody = 0x0, bcpc = 0x0, handlerstack = 0x6215a0, restartstack = 0x6215a0, prstack = 0x0, nodestack = 0x7ffff3b0f010, bcprottop = 0x7ffff3b0f010, srcref = 0x6215a0, browserfinish = 0, returnValue = 0x0, jumptarget = 0x0, jumpmask = 0}
        pargs = 0x2f92f78
        rho1 = 0x2f8d330
        pt = <optimized out>
        x = 0x92e890
        dots = <optimized out>
        nprotect = 3
#20 0x00007ffff7947a81 in R_DispatchOrEvalSP (ans=0x7fffffffbcc8, rho=0x6595c8, args=<optimized out>, generic=0x7ffff79d1669 "[", op=0x62a5e0, call=0x2f92df0) at subset.c:633
        prom = 0x2f92e98
        disp = <optimized out>
        prom = <optimized out>
        disp = <optimized out>
        x = <optimized out>
        il__x__ = <optimized out>
        irc__x__ = <optimized out>
        dl__x__ = <optimized out>
        drc__x__ = <optimized out>
        dl__x__ = <optimized out>
        drc__x__ = <optimized out>
#21 do_subset (call=0x2f92df0, op=0x62a5e0, args=0x2f92db8, rho=0x6595c8) at subset.c:653
        ans = 0x7ffff7856838 <Rf_findFun3+328>
#22 0x00007ffff78859e2 in Rf_eval (e=e@entry=0x2f92df0, rho=rho@entry=0x6595c8) at eval.c:798
        save = 2
        flag = 0
        vmax = 0x0
        op = 0x62a5e0
        tmp = <optimized out>
        evalcount = 508
        bcintactivesave = 0
        srcrefsave = 0x6215a0
        depthsave = 0
#23 0x00007ffff78b888a in Rf_ReplIteration (rho=0x6595c8, savestack=<optimized out>, browselevel=0, state=0x7fffffffbf80) at main.c:264
        c = <optimized out>
        browsevalue = <optimized out>
        value = <optimized out>
        wasDisplayed = FALSE
        thisExpr = 0x2f92df0
        state = <optimized out>
        savestack = <optimized out>
        rho = <optimized out>
        c = <optimized out>
        browsevalue = <optimized out>
        value = <optimized out>
        wasDisplayed = <optimized out>
        browselevel = <optimized out>
        thisExpr = <optimized out>
        c = <optimized out>
        browsevalue = <optimized out>
        value = <optimized out>
        thisExpr = <optimized out>
        wasDisplayed = FALSE
#24 0x00007ffff78b8c41 in R_ReplConsole (rho=0x6595c8, savestack=0, browselevel=0) at main.c:314
        status = <optimized out>
        state = {status = PARSE_OK, prompt_type = 1, browselevel = 0, buf = "library(data.table);M  <- 160;M4 <- M / 4;J  <- 10001;setDTthreads(threads = 80);d2 <- data.table(Time=rep(1:J, each=M),AgentID=rep(1:M4, times=4*J*M),i_phase=sample(0:1, J*M, replace=T)*2-1,State=sam"..., bufp = 0x7fffffffc1b6 ""}
#25 0x00007ffff78b8cd8 in run_Rmainloop () at main.c:1113
No locals.
#26 0x00007ffff78b8d32 in Rf_mainloop () at main.c:1120
No locals.
#27 0x00000000004007cb in main (ac=<optimized out>, av=<optimized out>) at Rmain.c:29
No locals.

But disabling checks for simple memory errors with

export MALLOC_CHECK_=3

before running R, this reveals now an error similar to the other ones:

Argument 'by' after substitute: .(Time, i_phase, N, SimulationID, discountFactor)
Detected that j uses these columns: [State]
Finding groups using forderv ... forder.c received 256025600 rows and 5 columns
15.2s elapsed (00:06:21 cpu) 
Finding group sizes from the positions (can be avoided to save RAM) ... 0.002s elapsed (0.011s cpu) 
Getting back original order ... forder.c received a vector type 'integer' length 80008

 *** caught segfault ***
address 0x7f02a5b1f000, cause 'invalid permissions'

Traceback:
 1: forderv(firstofeachgroup)
 2: `[.data.table`(d2, , `:=`(nnDistance_inphase, apply(as.matrix(dist(State,     method = "euclidean")), 1, FUN = findNNDistance)), by = .(Time,     i_phase, N, SimulationID, discountFactor), verbose = T)
 3: d2[, `:=`(nnDistance_inphase, apply(as.matrix(dist(State, method = "euclidean")),     1, FUN = findNNDistance)), by = .(Time, i_phase, N, SimulationID,     discountFactor), verbose = T]
An irrecoverable exception occurred. R is aborting now ...
Segmentation fault (core dumped)

And these errors - the invalid permission as well as the memory not mapped ones, all seem to originate from forder.c:995 in the radix_r() function.

The invalid permission ones seem to be related to the assignment of my_TMP there, as offset to be written is readable in gdb

(gdb) p my_TMP[my_starts[my_key[i]]+1]
$1 = 0
(gdb)
GDB Backtrace for `N = 80` 
#0  0x00007ffff058e9bd in radix_r (from=1640, to=<optimized out>, radix=1) at forder.c:995
        i = 75
        osub = 0x7fffb24f59e0
        my_starts = {0, 8, 8, 8, 8, 8, 16, 16, 16, 16, 16, 24, 24, 24, 24, 24, 32, 32, 32, 32, 32, 39, 40, 40, 40, 40, 48, 48, 48, 48, 48, 55, 56, 56, 56, 56, 64, 64, 64, 64, 64, 72, 72, 72, 72, 72, 
          76, 80, 80, 80, 80, 80, 88, 88, 88, 88, 88, 96, 96, 96, 96, 96, 103, 104, 104, 104, 104, 112, 112, 112, 112, 112, 120, 120, 120, 120, 120, 128, 128, 128, 128, 128, 136, 136, 136, 136, 136, 
          144, 144, 144, 144, 144, 152, 152, 152, 152, 152, 160, 160, 160, 160, 160, 168, 168, 168, 168, 168, 176, 176, 176, 176, 176, 183, 184, 184, 184, 184, 192, 192, 192, 192, 192, 199, 200, 200, 
          200, 200, 208, 208, 208, 208, 208, 216, 216, 216, 216, 216, 224, 224, 224, 224, 224, 231, 232, 232, 232, 232, 240, 240, 240, 240, 240, 248, 248, 248, 248, 248, 256, 256, 256, 256, 256, 264, 
          264, 264, 264, 264, 272, 272, 272, 272, 272, 280, 280, 280, 280, 280, 288, 288, 288, 288, 288, 296, 296, 296, 296, 296, 304, 304, 304, 304, 304, 311, 312, 312, 312, 312, 320, 320, 320...}
        my_starts_copy = {0, 0, 8, 8, 8, 8, 8, 16, 16, 16, 16, 16, 24, 24, 24, 24, 24, 32, 32, 32, 32, 32, 39, 40, 40, 40, 40, 48, 48, 48, 48, 48, 55, 56, 56, 56, 56, 64, 64, 64, 64, 64, 72, 72, 72, 
          72, 72, 80, 80, 80, 80, 80, 88, 88, 88, 88, 88, 96, 96, 96, 96, 96, 103, 104, 104, 104, 104, 112, 112, 112, 112, 112, 120, 120, 120, 120, 120, 128, 128, 128, 128, 128, 136, 136, 136, 136, 
          136, 144, 144, 144, 144, 144, 152, 152, 152, 152, 152, 160, 160, 160, 160, 160, 168, 168, 168, 168, 168, 176, 176, 176, 176, 176, 183, 184, 184, 184, 184, 192, 192, 192, 192, 192, 199, 200, 
          200, 200, 200, 208, 208, 208, 208, 208, 216, 216, 216, 216, 216, 224, 224, 224, 224, 224, 231, 232, 232, 232, 232, 240, 240, 240, 240, 240, 248, 248, 248, 248, 248, 256, 256, 256, 256, 256, 
          264, 264, 264, 264, 264, 272, 272, 272, 272, 272, 280, 280, 280, 280, 280, 288, 288, 288, 288, 288, 296, 296, 296, 296, 296, 304, 304, 304, 304, 304, 311, 312, 312, 312, 312, 320, 320, 320...}
        my_TMP = 0x7ffce85c6ed4
        my_counts = {0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 7, 1, 0, 0, 0, 8, 0, 0, 0, 0, 7, 1, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 
          0, 7, 1, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 7, 1, 0, 0, 0, 8, 0, 0, 0, 0, 7, 1, 0, 
          0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 7, 1, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 
          0, 0, 0, 7, 1, 0, 0, 0, 8, 0, 0, 0...}
        my_ugrp = 0x23da570 "f"
        my_key = 0x26f2848 "\001\001\001\001\001\001\001\001\006\006\006\006\006\006\006\006\v\v\v\v\v\v\v\v\020\020\020\020\020\020\020\020\025\025\025\026\025\025\025\025\032\032\032\032\032\032\032\032\037\037\037\037\037\037\037 $$$$$$$$))))))))........3333333388888888===>====BBBBBBBBGGGGGGGGLLLLLLLLQQQQQQQQVVVVVVVV[[[[[[[[````````eeeeeeeejjjjjjjjooopoooottttttttyyyyyyyz"...
        my_gs = 0x0
        ngrp = 58
        skip = <optimized out>
        my_n = 408
        batchSize = <optimized out>
        nBatch = <optimized out>
        lastBatchSize = <optimized out>
        counts = <optimized out>
        ugrps = <optimized out>
        ngrps = <optimized out>
        skip = <optimized out>
        n_rem = <optimized out>
        ugrp = "\000\000\b\000\b\000\b\000\b\000\b\000\020\000\020\000\020\000\020\000\020\000\030\000\030\000\030\000\030\000\030\000 \000 \000 \000 \000 \000'\000(\000(\000(\000(\000\060\000\060\000\060\000\060\000\060\000\067\000\070\000\070\000\070\000\070\000@\000@\000@\000@\000@\000H\000H\000H\000H\000H\000L\000P\000P\000P\000P\000P\000X\000X\000X\000X\000X\000`\000`\000`\000`\000`\000g\000h\000h\000h\000h\000p\000p\000p\000p\000p\000x\000x\000x\000x\000x\000\200\000\200\000\200\000\200\000\200\000\210\000\210\000\210\000\210\000\210\000\220\000\220\000\220\000\220\000\220\000\230\000\230\000\230\000\230\000\230\000\240\000\240\000\240\000"...
        seen = {false, false, false, false, 8, false, 8, false, 8, false, 8, false, 8, false, 16, false, 16, false, 16, false, 16, false, 16, false, 24, false, 24, false, 24, false, 24, false, 24, 
          false, 32, false, 32, false, 32, false, 32, false, 32, false, 39, false, 40, false, 40, false, 40, false, 40, false, 48, false, 48, false, 48, false, 48, false, 48, false, 55, false, 56, 
          false, 56, false, 56, false, 56, false, 64, false, 64, false, 64, false, 64, false, 64, false, 72, false, 72, false, 72, false, 72, false, 72, false, 80, false, 80, false, 80, false, 80, 
          false, 80, false, 88, false, 88, false, 88, false, 88, false, 88, false, 96, false, 96, false, 96, false, 96, false, 96, false, 103, false, 104, false, 104, false, 104, false, 104, false, 
          112, false, 112, false, 112, false, 112, false, 112, false, 120, false, 120, false, 120, false, 120, false, 120, false, 128, false, 128, false, 128, false, 128, false, 128, false, 136, false, 
          136, false, 136, false, 136, false, 136, false, 144, false, 144, false, 144, false, 144, false, 144, false, 152, false, 152, false, 152, false, 152, false, 152, false, 160, false, 160, false, 
          160, false...}
        ngrp = <optimized out>
        last_seen = <optimized out>
        starts = <optimized out>
        my_gs = 0x0
#1  0x00007ffff0590828 in radix_r (from=<optimized out>, to=<optimized out>, radix=<optimized out>) at forder.c:1238
        start = <optimized out>
        start = <optimized out>
        i = <optimized out>
        i = <optimized out>
        anyBig = <optimized out>
        my_n = 0
        batchSize = <optimized out>
        nBatch = <optimized out>
        lastBatchSize = <optimized out>
        counts = <optimized out>
        ugrps = <optimized out>
        ngrps = <optimized out>
        skip = <optimized out>
        n_rem = <optimized out>
        ugrp = "X\000`\000`\000`\000`\000`\000g\000h\000h\000h\000h\000p\000p\000p\000p\000p\000x\000x\000x\000x\000x\000\200\000\200\000\200\000\200\000\200\000\210\000\210\000\210\000\210\000\210\000\220\000\220\000\220\000\220\000\220\000\230\000\230\000\230\000\230\000\230\000\240\000\240\000\240\000\240\000\240\000\250\000\250\000\250\000\250\000\250\000\260\000\260\000\260\000\260\000\260\000\267\000\270\000\270\000\270\000\270\000\300\000\300\000\300\000\300\000\300\000\307\000\310\000\310\000\310\000\310\000\320\000\320\000\320\000\320\000\320\000\330\000\330\000\330\000\330\000\330\000\340\000\340\000\340\000\340\000\340\000\347\000\350\000\350\000\350\000\350\000\360\000\360\000\360\000\360\000\360\000\370\000\370\000\370\000\370\000"...
        seen = {88, false, 96, false, 96, false, 96, false, 96, false, 96, false, 103, false, 104, false, 104, false, 104, false, 104, false, 112, false, 112, false, 112, false, 112, false, 112, false, 120, false, 120, false, 120, false, 120, false, 120, false, 128, false, 128, false, 128, false, 128, false, 128, false, 136, false, 136, false, 136, false, 136, false, 136, false, 144, false, 144, false, 144, false, 144, false, 144, false, 152, false, 152, false, 152, false, 152, false, 152, false, 160, false, 160, false, 160, false, 160, false, 160, false, 168, false, 168, false, 168, false, 168, false, 168, false, 176, false, 176, false, 176, false, 176, false, 176, false, 183, false, 184, false, 184, false, 184, false, 184, false, 192, false, 192, false, 192, false, 192, false, 192, false, 199, false, 200, false, 200, false, 200, false, 200, false, 208, false, 208, false, 208, false, 208, false, 208, false, 216, false, 216, false, 216, false, 216, false, 216, false, 224, false, 224, false, 224, false, 224, false, 224, false, 231, false, 232, false, 232, false, 232, false, 232, false, 240, false, 240, false, 240, false, 240, false, 240, false, 248, false, 248, false, 248, false, 248, false...}
        ngrp = <optimized out>
        last_seen = <optimized out>
        starts = <optimized out>
        my_gs = 0x0
#2  0x00007ffff72eb46e in ?? () from /usr/lib64/libgomp.so.1
No symbol table info available.
#3  0x00007ffff70b94f9 in start_thread () from /lib64/libpthread.so.0
No symbol table info available.
#4  0x00007ffff6df1ecf in clone () from /lib64/libc.so.6
No symbol table info available.

But the memory not mapped look like out of bounds read access to my_TMP, as the offet is not readable in gdb

(gdb) p my_TMP[my_starts[my_key[i]]+1]
Cannot access memory at address 0x4a8405c
(gdb) p my_starts[my_key[i]]
$1 = 1
(gdb) p osub[i]
$2 = 23761
(gdb)
GDB Backtrace for `N = 96` 
#0  0x00007ffff05f19bd in radix_r (from=23760, to=<optimized out>, radix=1) at forder.c:995
        i = 0
        osub = 0x358d2b0
        my_starts = {0, 0, 1, 8, 8, 8, 8, 8, 16, 16, 16, 16, 16, 24, 24, 24, 24, 24, 32, 32, 32, 32, 32, 40, 40, 40, 40, 40, 47, 48, 48, 48, 48, 55, 56, 56, 56, 56, 64, 64, 64, 64, 64, 72, 72, 72, 72, 
          72, 80, 80, 80, 80, 80, 88, 88, 88, 88, 88, 96, 96, 96, 96, 96, 104, 104, 104, 104, 104, 111, 112, 112, 112, 112, 120, 120, 120, 120, 120, 127, 128, 128, 128, 128, 136, 136, 136, 136, 136, 
          143, 144, 144, 144, 144, 152, 152, 152, 152, 152, 160, 160, 160, 160, 160, 167, 168, 168, 168, 168, 176, 176, 176, 176, 176, 184, 184, 184, 184, 184, 192, 192, 192, 192, 192, 199, 200, 200, 
          200, 200, 208, 208, 208, 208, 208, 216, 216, 216, 216, 216, 224, 224, 224, 224, 224, 232, 232, 232, 232, 232, 240, 240, 240, 240, 240, 248, 248, 248, 248, 248, 256, 256, 256, 256, 256, 264, 
          264, 264, 264, 264, 272, 272, 272, 272, 272, 279, 280, 280, 280, 280, 288, 288, 288, 288, 288, 296, 296, 296, 296, 296, 304, 304, 304, 304, 304, 312, 312, 312, 312, 312, 320, 320...}
        my_starts_copy = {0, 0, 0, 8, 8, 8, 8, 8, 16, 16, 16, 16, 16, 24, 24, 24, 24, 24, 32, 32, 32, 32, 32, 40, 40, 40, 40, 40, 47, 48, 48, 48, 48, 55, 56, 56, 56, 56, 64, 64, 64, 64, 64, 72, 72, 72, 
          72, 72, 80, 80, 80, 80, 80, 88, 88, 88, 88, 88, 96, 96, 96, 96, 96, 104, 104, 104, 104, 104, 111, 112, 112, 112, 112, 120, 120, 120, 120, 120, 127, 128, 128, 128, 128, 136, 136, 136, 136, 
          136, 143, 144, 144, 144, 144, 152, 152, 152, 152, 152, 160, 160, 160, 160, 160, 167, 168, 168, 168, 168, 176, 176, 176, 176, 176, 184, 184, 184, 184, 184, 192, 192, 192, 192, 192, 199, 200, 
          200, 200, 200, 208, 208, 208, 208, 208, 216, 216, 216, 216, 216, 224, 224, 224, 224, 224, 232, 232, 232, 232, 232, 240, 240, 240, 240, 240, 248, 248, 248, 248, 248, 256, 256, 256, 256, 256, 
          264, 264, 264, 264, 264, 272, 272, 272, 272, 272, 279, 280, 280, 280, 280, 288, 288, 288, 288, 288, 296, 296, 296, 296, 296, 304, 304, 304, 304, 304, 312, 312, 312, 312, 312, 320, 320...}
        my_TMP = 0x4a84054
        my_counts = {0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 7, 1, 0, 0, 0, 7, 1, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 
          0, 0, 8, 0, 0, 0, 0, 7, 1, 0, 0, 0, 8, 0, 0, 0, 0, 7, 1, 0, 0, 0, 8, 0, 0, 0, 0, 7, 1, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 7, 1, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 7, 1, 
          0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 7, 1, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 
          0, 0, 0, 0, 8, 0, 0, 0, 0, 8, 0, 0...}
        my_ugrp = 0x2adea70 "\200"
        my_key = 0x3176240 "\002\002\002\002\002\002\002\002\a\a\a\a\a\a\a\a\f\f\f\f\f\f\f\f\021\021\021\021\021\021\021\021\026\026\026\026\026\026\026\026\033\033\033\033\033\033\033\034       !%%%%%%%%********////////4444444499999999>>>>>>>>CCCCCCCDHHHHHHHHMMMMMMMNRRRRRRRRWWWWWWWX\\\\\\\\\\\\\\\\aaaaaaaafffgffffkkkkkkkkppppppppuuuuuuuuzzzzzzz{"...
        my_gs = 0x0
        ngrp = 60
        skip = <optimized out>
        my_n = 408
        batchSize = <optimized out>
        nBatch = <optimized out>
        lastBatchSize = <optimized out>
        counts = <optimized out>
        ugrps = <optimized out>
        ngrps = <optimized out>
        skip = <optimized out>
        n_rem = <optimized out>
        ugrp = "\000\000\000\000\001\000\b\000\b\000\b\000\b\000\b\000\020\000\020\000\020\000\020\000\020\000\030\000\030\000\030\000\030\000\030\000 \000 \000 \000 \000 \000(\000(\000(\000(\000(\000/\000\060\000\060\000\060\000\060\000\067\000\070\000\070\000\070\000\070\000@\000@\000@\000@\000@\000H\000H\000H\000H\000H\000P\000P\000P\000P\000P\000X\000X\000X\000X\000X\000`\000`\000`\000`\000`\000h\000h\000h\000h\000h\000o\000p\000p\000p\000p\000x\000x\000x\000x\000x\000\177\000\200\000\200\000\200\000\200\000\210\000\210\000\210\000\210\000\210\000\217\000\220\000\220\000\220\000\220\000\230\000\230\000\230\000\230\000\230\000\240\000\240\000"...
        seen = {false, false, false, false, false, false, 8, false, 8, false, 8, false, 8, false, 8, false, 16, false, 16, false, 16, false, 16, false, 16, false, 24, false, 24, false, 24, false, 24, 
          false, 24, false, 32, false, 32, false, 32, false, 32, false, 32, false, 40, false, 40, false, 40, false, 40, false, 40, false, 47, false, 48, false, 48, false, 48, false, 48, false, 55, 
          false, 56, false, 56, false, 56, false, 56, false, 64, false, 64, false, 64, false, 64, false, 64, false, 72, false, 72, false, 72, false, 72, false, 72, false, 80, false, 80, false, 80, 
          false, 80, false, 80, false, 88, false, 88, false, 88, false, 88, false, 88, false, 96, false, 96, false, 96, false, 96, false, 96, false, 104, false, 104, false, 104, false, 104, false, 104, 
          false, 111, false, 112, false, 112, false, 112, false, 112, false, 120, false, 120, false, 120, false, 120, false, 120, false, 127, false, 128, false, 128, false, 128, false, 128, false, 136, 
          false, 136, false, 136, false, 136, false, 136, false, 143, false, 144, false, 144, false, 144, false, 144, false, 152, false, 152, false, 152, false, 152, false, 152, false, 160, false, 160, 
          false...}
        ngrp = <optimized out>
        last_seen = <optimized out>
        starts = <optimized out>
        my_gs = 0x0
#1  0x00007ffff05f3828 in radix_r (from=<optimized out>, to=<optimized out>, radix=<optimized out>) at forder.c:1238
        start = <optimized out>
        start = <optimized out>
        i = <optimized out>
        i = <optimized out>
        anyBig = <optimized out>
        my_n = 0
        batchSize = <optimized out>
        nBatch = <optimized out>
        lastBatchSize = <optimized out>
        counts = <optimized out>
        ugrps = <optimized out>
        ngrps = <optimized out>
        skip = <optimized out>
        n_rem = <optimized out>
        ugrp = "X\000X\000`\000`\000`\000`\000`\000h\000h\000h\000h\000h\000o\000p\000p\000p\000p\000x\000x\000x\000x\000x\000\177\000\200\000\200\000\200\000\200\000\210\000\210\000\210\000\210\000\210\000\217\000\220\000\220\000\220\000\220\000\230\000\230\000\230\000\230\000\230\000\240\000\240\000\240\000\240\000\240\000\247\000\250\000\250\000\250\000\250\000\260\000\260\000\260\000\260\000\260\000\270\000\270\000\270\000\270\000\270\000\300\000\300\000\300\000\300\000\300\000\307\000\310\000\310\000\310\000\310\000\320\000\320\000\320\000\320\000\320\000\330\000\330\000\330\000\330\000\330\000\340\000\340\000\340\000\340\000\340\000\350\000\350\000\350\000\350\000\350\000\360\000\360\000\360\000\360\000\360\000\370\000\370\000\370\000"...
        seen = {88, false, 88, false, 96, false, 96, false, 96, false, 96, false, 96, false, 104, false, 104, false, 104, false, 104, false, 104, false, 111, false, 112, false, 112, false, 112, false, 112, false, 120, false, 120, false, 120, false, 120, false, 120, false, 127, false, 128, false, 128, false, 128, false, 128, false, 136, false, 136, false, 136, false, 136, false, 136, false, 143, false, 144, false, 144, false, 144, false, 144, false, 152, false, 152, false, 152, false, 152, false, 152, false, 160, false, 160, false, 160, false, 160, false, 160, false, 167, false, 168, false, 168, false, 168, false, 168, false, 176, false, 176, false, 176, false, 176, false, 176, false, 184, false, 184, false, 184, false, 184, false, 184, false, 192, false, 192, false, 192, false, 192, false, 192, false, 199, false, 200, false, 200, false, 200, false, 200, false, 208, false, 208, false, 208, false, 208, false, 208, false, 216, false, 216, false, 216, false, 216, false, 216, false, 224, false, 224, false, 224, false, 224, false, 224, false, 232, false, 232, false, 232, false, 232, false, 232, false, 240, false, 240, false, 240, false, 240, false, 240, false, 248, false, 248, false, 248, false...}
        ngrp = <optimized out>
        last_seen = <optimized out>
        starts = <optimized out>
        my_gs = 0x0
#2  0x00007ffff72eb46e in ?? () from /usr/lib64/libgomp.so.1
No symbol table info available.
#3  0x00007ffff70b94f9 in start_thread () from /lib64/libpthread.so.0
No symbol table info available.
#4  0x00007ffff6df1ecf in clone () from /lib64/libc.so.6
No symbol table info available.

There was a part here about some address calculation, which turned out to be wrong because I forgot someting in the calulation. And doing the calulation right makes the part irrelevant. But I'll leave it as Spoiler for completeness sake.

Wrong asumption

What is also interesting doing some address calculation to track this down. my_TMP is assigned priviously in forder.c:987 with

int *restrict my_TMP = TMP + omp_get_thread_num()*UINT16_MAX;

when doing this calulation backwards with the address found by gdb

(gdb) p my_TMP
$3 = (int * restrict) 0x4a84054
(gdb) p TMP
$4 = (int *) 0x35c41a0
(gdb)

it turns out (4a84054−35c41a0)/ffff = 14bfeb4/ffff = 14c which is 332 decimal which would mean omp_get_thread_num() would have been 332. Which is way more than the 96 threads asked for, as per OpenMP documentation this should be smaller than the current number of running thread for that parallel region.

Also it look like with MALLOC_CHECK_=3 the memory not mapped error becomes an invalid permission error which would suggest that the invalid permission ones are root of the problem. Running the memory calculation with these ones

(gdb) p my_TMP
$1 = (int * restrict) 0x7ffce85c6ed4
(gdb) p TMP
$2 = (int *) 0x7ffce7207010
(gdb)

tell a similar story: (7ffce85c6ed4−7ffce7207010)/ffff = 13c which is 316. So something looks to be off with the number of theads started which causes some out of bounds access for the gobal TMP buffer. But I still have no idea to why that happens

bennet0496 added a commit to bennet0496/data.table that referenced this issue Jul 28, 2021
@bennet0496
fixing "invalid permission" segfault from Rdatatable#5077 by only sta…
3dd68c9 
…rting the amount of thread, memory was allocated for.

L717-L718 allocated memory for a batch-throttled number of threads in the TMP buffer, but L1235-L1238 started an unthrottled number of threads, causing the segfault at L995 for threads over the throttled threshold.
@mattdowle mattdowle added this to the 1.14.1 milestone Jul 29, 2021
@mattdowle mattdowle added the bug label Jul 29, 2021
@mattdowle mattdowle mentioned this issue Jul 29, 2021
Merged
@mattdowle mattdowle added the High label Jul 29, 2021
bennet0496 added a commit to bennet0496/data.table that referenced this issue Jul 30, 2021
@bennet0496
updating memory allocation for global TMP and adding safe-guard to no…
2b8c12e 
…t start too many threads where the buffer might be used. Also renaming the global TMP as such. see Rdatatable#5077 and Rdatatable#5087
mattdowle pushed a commit that referenced this issue Aug 4, 2021
@bennet0496
fixing "invalid permission" segfault from #5077 by only starting the …
6452c70 
…amount of thread, memory was allocated for. (#5087)
@ben-schwen ben-schwen mentioned this issue Aug 5, 2021
Closed
@MichaelChirico MichaelChirico mentioned this issue Sep 21, 2021
Closed
@ben-schwen ben-schwen mentioned this issue Sep 30, 2021
Closed
@ben-schwen ben-schwen mentioned this issue Sep 28, 2022
Open
@bschilder bschilder mentioned this issue Mar 21, 2023
Closed
@jangorecki jangorecki modified the milestones: 1.14.9, 1.15.0 Oct 29, 2023
@ben-schwen ben-schwen mentioned this issue Nov 8, 2023
Closed
@j3551ca j3551ca mentioned this issue Jul 30, 2024
Closed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
bug High
Projects
None yet
Milestone
1.15.0
3 participants
@mattdowle @jangorecki @bennet0496