rpms/ksh
7
0
Fork 0
Code Issues Pull Requests Releases Activity

Resolves: RHEL-20345 - Fix crash on failure to trim ~/.sh_history

Browse Source
This commit is contained in:
Vincent Mihalkovic 2024-01-03 15:19:39 +01:00
parent 34d2578507
commit 8b51f18f60
2 changed files with 106 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

99
ksh-1.0.7-history-trim.patch Normal file
View File

@ -0,0 +1,99 @@
From 2075b2b96208ac8b989ca316dcdd674c3f488e2b Mon Sep 17 00:00:00 2001
From: Martijn Dekker <martijn@inlv.org>
Date: Thu, 28 Dec 2023 04:02:28 +0000
Subject: [PATCH] Fix crash on failure to trim ~/.sh_history
@vmihalko writes:
> We were able to reproduce an old issue mentioned in
> https://bugzilla.redhat.com/show_bug.cgi?id=1885399 using the
> latest version of ksh. The corresponding code has not changed
> much in the past few years.
>
> To provide further explanation, the problem arises when a user's
> .sh_history file grows to a size that triggers the hist_trim
> function, but the user lacks (after the creation of .sh_history)
> the necessary write permissions to their $HOME directory. As a
> result, ksh becomes stuck in a recursive loop between the
> sh_histinit(src/cmd/ksh93/edit/history.c#L203) function and the
> hist_trim(src/cmd/ksh93/edit/history.c#L417) function.
>
> Conditions for reproduction:
>
> 1. The size of the .sh_history file is larger than the HIST_MAX
> limit. (src/cmd/ksh93/edit/history.c, line 325)
> 2. .sh_history file has not been changed in the HIST_RECENT
> seconds (src/cmd/ksh93/edit/history.c, line 406)
> 3. The user does not have permission to write to the $HOME
> directory.
src/cmd/ksh93/edit/history.c: hist_trim():
- Print a warning and return if unlink(2) fails. The warning tells
the user to check the history file's parent directory is
writable. This is the best I realistically do for now, because
this function's basic method assumes a writable parent directory.
- The temp file fallback is deleted because it's fundamentally
flawed: it assumes the temp file is made on the same volume as
the history file and can simply be rename(2)'d in place. Even
on systems where this is the case, it doesn't appear to be
working correctly, but this is not worth looking into.
Resolves: https://github.com/ksh93/ksh/issues/695
---
src/cmd/ksh93/edit/history.c | 34 ++++------------------------------
1 files changed, 4 insertions(+), 30 deletions(-)
diff --git a/src/cmd/ksh93/edit/history.c b/src/cmd/ksh93/edit/history.c
index c49c23244d00..f3672bf6e220 100644
--- a/src/cmd/ksh93/edit/history.c
+++ b/src/cmd/ksh93/edit/history.c
@@ -419,34 +419,13 @@ static History_t* hist_trim(History_t *hp, int n)
char *cp;
int incmd=1, c=0;
History_t *hist_new, *hist_old = hp;
- char *buff, *endbuff, *tmpname=0;
+ char *buff, *endbuff;
off_t oldp,newp;
struct stat statb;
- unlink(hist_old->histname);
- if(access(hist_old->histname,F_OK) >= 0)
+ if(unlink(hist_old->histname) < 0)
{
- /* The unlink can fail on Windows 95 */
- int fd;
- char *last, *name=hist_old->histname;
- sh_close(sffileno(hist_old->histfp));
- tmpname = (char*)sh_malloc(strlen(name)+14);
- if(last = strrchr(name,'/'))
- {
- *last = 0;
- pathtmp(tmpname,name,"hist",NULL);
- *last = '/';
- }
- else
- pathtmp(tmpname,e_dot,"hist",NULL);
- if(rename(name,tmpname) < 0)
- {
- free(tmpname);
- tmpname = name;
- }
- fd = open(tmpname,O_RDONLY|O_cloexec);
- sfsetfd(hist_old->histfp,fd);
- if(tmpname==name)
- tmpname = 0;
+ errormsg(SH_DICT,ERROR_warn(0),"cannot trim history file %s; make sure parent directory is writable",hist_old->histname);
+ return hist_ptr = hist_old;
}
hist_ptr = 0;
if(fstat(sffileno(hist_old->histfp),&statb)>=0)
@@ -501,11 +480,6 @@ static History_t* hist_trim(History_t *hp, int n)
}
hist_cancel(hist_new);
sfclose(hist_old->histfp);
- if(tmpname)
- {
- unlink(tmpname);
- free(tmpname);
- }
free((char*)hist_old);
return hist_ptr = hist_new;
}

8
ksh.spec
View File

@ -4,7 +4,7 @@ URL: http://www.kornshell.com/
License: EPL-1.0 License: EPL-1.0
Epoch: 3 Epoch: 3
Version: 1.0.6 Version: 1.0.6
Release: 1%{?dist} Release: 2%{?dist}
Source0: https://github.com/ksh93/%{name}/archive/v%{version}/%{name}-%{version}.tar.gz Source0: https://github.com/ksh93/%{name}/archive/v%{version}/%{name}-%{version}.tar.gz
Source1: kshcomp.conf Source1: kshcomp.conf
Source2: kshrc.rhs Source2: kshrc.rhs
@ -20,6 +20,9 @@ Patch1: %{name}-1.0.6-alarm-1.patch
# alarm fixing patch from https://github.com/ksh93/ksh/issues/422#issuecomment-1581168550 # alarm fixing patch from https://github.com/ksh93/ksh/issues/422#issuecomment-1581168550
Patch2: %{name}-1.0.6-alarm-2.patch Patch2: %{name}-1.0.6-alarm-2.patch
#https://github.com/ksh93/ksh/commit/2075b2b96208ac8b989ca316dcdd674c3f488e2b
Patch3: %{name}-1.0.7-history-trim.patch
Conflicts: pdksh Conflicts: pdksh
Requires: coreutils, diffutils Requires: coreutils, diffutils
BuildRequires: gcc BuildRequires: gcc
@ -141,6 +144,9 @@ fi
%config(noreplace) %{_sysconfdir}/binfmt.d/kshcomp.conf %config(noreplace) %{_sysconfdir}/binfmt.d/kshcomp.conf
%changelog %changelog
* Wed Jan 03 2024 Vincent Mihalkovic <vmihalko@redhat.com> - 3:1.0.6-2
- Fix crash on failure to trim ~/.sh_history (#20345)
* Mon Sep 18 2023 Vincent Mihalkovic <vmihalko@redhat.com> - 3:1.0.6-1 * Mon Sep 18 2023 Vincent Mihalkovic <vmihalko@redhat.com> - 3:1.0.6-1
- Rebase to non-beta version, because upstream says it should to be used - Rebase to non-beta version, because upstream says it should to be used
https://github.com/ksh93/ksh/issues/667#issuecomment-1653665697 https://github.com/ksh93/ksh/issues/667#issuecomment-1653665697